Chapter 9 The Nature of Predation Cap 9.pdf · rate at which the adult females lay eggs. Each egg...

••

9.1 Introduction: the types of predators

Consumers affect the distribution and abundance of the thingsthey consume and vice versa, and these effects are of central impor-tance in ecology. Yet, it is never an easy task to determine whatthe effects are, how they vary and why they vary. These topicswill be dealt with in this and the next few chapters. We beginhere by asking ‘What is the nature of predation?’, ‘What are theeffects of predation on the predators themselves and on their prey?’and ‘What determines where predators feed and what they feedon?’ In Chapter 10, we turn to the consequences of predation forthe dynamics of predator and prey populations.

Predation, put simply, is consumptionof one organism (the prey) by anotherorganism (the predator), in which theprey is alive when the predator first

attacks it. This excludes detritivory, the consumption of deadorganic matter, which is discussed in its own right in Chapter 11.Nevertheless, it is a definition that encompasses a wide varietyof interactions and a wide variety of ‘predators’.

There are two main ways in whichpredators can be classified. Neither isperfect, but both can be useful. Themost obvious classification is ‘taxo-nomic’: carnivores consume animals,herbivores consume plants and omni-

vores consume both (or, more correctly, prey from more thanone trophic level – plants and herbivores, or herbivores and carnivores). An alternative, however, is a ‘functional’ classificationof the type already outlined in Chapter 3. Here, there are fourmain types of predator: true predators, grazers, parasitoids andparasites (the last is divisible further into microparasites and macro-

parasites as explained in Chapter 12).True predators kill their prey more

or less immediately after attacking

them; during their lifetime they kill several or many different preyindividuals, often consuming prey in their entirety. Most of themore obvious carnivores like tigers, eagles, coccinellid beetles andcarnivorous plants are true predators, but so too are seed-eatingrodents and ants, plankton-consuming whales, and so on.

Grazers also attack large numbers ofprey during their lifetime, but theyremove only part of each prey individ-ual rather than the whole. Their effect on a prey individual,although typically harmful, is rarely lethal in the short term, andcertainly never predictably lethal (in which case they would betrue predators). Amongst the more obvious examples are the largevertebrate herbivores like sheep and cattle, but the flies that bitea succession of vertebrate prey, and leeches that suck theirblood, are also undoubtedly grazers by this definition.

Parasites, like grazers, consume partsof their prey (their ‘host’), rather thanthe whole, and are typically harmful butrarely lethal in the short term. Unlike grazers, however, their attacks are concentrated on one or a very few individuals duringtheir life. There is, therefore, an intimacy of association betweenparasites and their hosts that is not seen in true predators and grazers. Tapeworms, liver flukes, the measles virus, the tuberculosisbacterium and the flies and wasps that form mines and galls onplants are all obvious examples of parasites. There are also manyplants, fungi and microorganisms that are parasitic on plants(often called ‘plant pathogens’), including the tobacco mosaic virus, the rusts and smuts and the mistletoes. Moreover, manyherbivores may readily be thought of as parasites. For example,aphids extract sap from one or a very few individual plants with which they enter into intimate contact. Even caterpillars oftenrely on a single plant for their development. Plant pathogens, and animals parasitic on animals, will be dealt with together inChapter 12. ‘Parasitic’ herbivores, like aphids and caterpillars, aredealt with here and in the next chapter, where we group them

definition of

predation

taxonomic and

functional

classifications

of predators

true predators

grazers

parasites

Chapter 9

The Nature of Predation

EIPC09 10/24/05 2:01 PM Page 266

THE NATURE OF PREDATION 267

together with true predators, grazers and parasitoids under theumbrella term ‘predator’.

The parasitoids are a group ofinsects that belong mainly to the orderHymenoptera, but also include many

Diptera. They are free-living as adults, but lay their eggs in, onor near other insects (or, more rarely, in spiders or woodlice). Thelarval parasitoid then develops inside or on its host. Initially, itdoes little apparent harm, but eventually it almost totally consumesthe host and therefore kills it. An adult parasitoid emerges fromwhat is apparently a developing host. Often, just one parasitoiddevelops from each host, but in some cases several or many indi-viduals share a host. Thus, parasitoids are intimately associatedwith a single host individual (like parasites), they do not causeimmediate death of the host (like parasites and grazers), but theireventual lethality is inevitable (like predators). For parasitoids, andalso for the many herbivorous insects that feed as larvae onplants, the rate of ‘predation’ is determined very largely by therate at which the adult females lay eggs. Each egg is an ‘attack’on the prey or host, even though it is the larva that hatches fromthe egg that does the eating.

Parasitoids might seem to be an unusual group of limited general importance. However, it has been estimated that theyaccount for 10% or more of the world’s species (Godfray, 1994).This is not surprising given that there are so many species of insects,that most of these are attacked by at least one parasitoid, and thatparasitoids may in turn be attacked by parasitoids. A number ofparasitoid species have been intensively studied by ecologists, andthey have provided a wealth of information relevant to predationgenerally.

In the remainder of this chapter, we examine the nature ofpredation. We will look at the effects of predation on the preyindividual (Section 9.2), the effects on the prey population as awhole (Section 9.3) and the effects on the predator itself (Section9.4). In the cases of attacks by true predators and parasitoids, theeffects on prey individuals are very straightforward: the prey iskilled. Attention will therefore be placed in Section 9.2 on preysubject to grazing and parasitic attack, and herbivory will be theprincipal focus. Apart from being important in its own right, her-bivory serves as a useful vehicle for discussing the subtleties andvariations in the effects that predators can have on their prey.

Later in the chapter we turn our attention to the behavior ofpredators and discuss the factors that determine diet (Section 9.5)and where and when predators forage (Section 9.6). These topicsare of particular interest in two broad contexts. First, foraging is an aspect of animal behavior that is subject to the scrutiny ofevolutionary biologists, within the general field of ‘behavioral ecology’. The aim, put simply, is to try to understand how naturalselection has favored particular patterns of behavior in particularcircumstances (how, behaviorally, organisms match their envir-onment). Second, the various aspects of predatory behavior canbe seen as components that combine to influence the population

dynamics of both the predator itself and its prey. The populationecology of predation is dealt with much more fully in the nextchapter.

9.2 Herbivory and individual plants: toleranceor defense

The effects of herbivory on a plant depend on which herbivoresare involved, which plant parts are affected, and the timing of attack relative to the plant’s development. In some insect–plantinteractions as much as 140 g, and in others as little as 3 g, of planttissue are required to produce 1 g of insect tissue (Gavloski & Lamb,2000a) – clearly some herbivores will have a greater impact thanothers. Moreover, leaf biting, sap sucking, mining, flower and fruitdamage and root pruning are all likely to differ in the effect theyhave on the plant. Furthermore, the consequences of defoliatinga germinating seedling are unlikely to be the same as those ofdefoliating a plant that is setting its own seed. Because the plantusually remains alive in the short term, the effects of herbivoryare also crucially dependent on the response of the plant. Plants may show tolerance of herbivore damage or resistance to attack.

9.2.1 Tolerance and plant compensation

Plant compensation is a term thatrefers to the degree of tolerance exhib-ited by plants. If damaged plants havegreater fitness than their undamagedcounterparts, they have overcompensated, and if they have lowerfitness, they have undercompensated for herbivory (Strauss &Agrawal, 1999). Individual plants can compensate for the effectsof herbivory in a variety of ways. In the first place, the removalof shaded leaves (with their normal rates of respiration but lowrates of photosynthesis; see Chapter 3) may improve the balancebetween photosynthesis and respiration in the plant as a whole.Second, in the immediate aftermath of an attack from a herbi-vore, many plants compensate by utilizing reserves stored in avariety of tissues and organs or by altering the distribution of photosynthate within the plant. Herbivore damage may alsolead to an increase in the rate of photosynthesis per unit area ofsurviving leaf. Often, there is compensatory regrowth of defoli-ated plants when buds that would otherwise remain dormant arestimulated to develop. There is also, commonly, a reduced deathrate of surviving plant parts. Clearly, then, there are a numberof ways in which individual plants compensate for the effects ofherbivory (discussed further in Sections 9.2.3–9.2.5). But perfectcompensation is rare. Plants are usually harmed by herbivores even though the compensatory reactions tend to counteract theharmful effects.

••

parasitoids

individual plants can

compensate for

herbivore effects

EIPC09 10/24/05 2:01 PM 67

268 CHAPTER 9

9.2.2 Defensive responses of plants

The evolutionary selection pressureexerted by herbivores has led to a variety of plant physical and chemicaldefenses that resist attack (see Sections

3.7.3 and 3.7.4). These may be present and effective continuously(constitutive defense) or increased production may be induced byattack (inducible defence) (Karban et al., 1999). Thus, production ofthe defensive hydroxamic acid is induced when aphids (Rhopalo-siphum padi) attack the wild wheat Triticum uniaristatum (Gianoli& Niemeyer, 1997), and the prickles of dewberries on cattle-grazedplants are longer and sharper than those on ungrazed plantsnearby (Abrahamson, 1975). Particular attention has been paid to rapidly inducible defenses, often the production of chemicalswithin the plant that inhibit the protease enzymes of the herbi-vores. These changes can occur within individual leaves, withinbranches or throughout whole tree canopies, and they may bedetectable within a few hours, days or weeks, and last a few days,weeks or years; such responses have now been reported in morethan 100 plant–herbivore systems (Karban & Baldwin, 1997).

There are, however, a number ofproblems in interpreting these responses(Schultz, 1988). First, are they ‘responses’

at all, or merely an incidental consequence of regrowth tissue having different properties from that removed by the herbivores? In fact, this issue is mainly one of semantics – if the metabolicresponses of a plant to tissue removal happen to be defensive, then natural selection will favor them and reinforce their use. Afurther problem is much more substantial: are induced chemicalsactually defensive in the sense of having an ecologically significanteffect on the herbivores that seem to have induced them? Finally,and of most significance, are they truly defensive in the sense ofhaving a measurable, positive impact on the plant making them,especially after the costs of mounting the response have been takeninto account?

Fowler and Lawton (1985) ad-dressed the second problem – ‘are theresponses harmful to the herbivores?’ – by reviewing the effects of rapidlyinducible plant defenses and found

little clear-cut evidence that they are effective against insect herbivores, despite a widespread belief that they were. Forexample, they found that most laboratory studies revealed onlysmall adverse effects (less than 11%) on such characters as larvaldevelopment time and pupal weight, with many studies thatclaimed a larger effect being flawed statistically, and they arguedthat such effects may have negligible consequences for field populations. However, there are also a number of cases, manyof which have been published since Fowler and Lawton’sreview, in which the plant’s responses do seem to be genuinelyharmful to the herbivores. When larch trees were defoliated by

the larch budmoth, Zeiraphera diniana, the survival and adultfecundity of the moths were reduced throughout the succeeding4–5 years as a combined result of delayed leaf production, tougherleaves, higher fiber and resin concentration and lower nitrogenlevels (Baltensweiler et al., 1977). Another common response toleaf damage is early abscission (‘dropping off ’) of mined leaves;in the case of the leaf-mining insect Phyllonorycter spp. on willowtrees (Salix lasiolepis), early abscission of mined leaves was an important mortality factor for the moths – that is, the herbivoreswere harmed by the response (Preszler & Price, 1993). As a final example, a few weeks of grazing on the brown seaweedAscophyllum nodosum by snails (Littorina obtusata) induces sub-stantially increased concentrations of phlorotannins (Figure 9.1a),which reduce further snail grazing (Figure 9.1b). In this case, simple clipping of the plants did not have the same effect as theherbivore. Indeed, grazing by another herbivore, the isopod Idoteagranulosa, also failed to induce the chemical defense. The snails canstay and feed on the same plant for long time periods (the isopodsare much more mobile), so that induced responses that take timeto develop can still be effective in reducing damage by snails.

The final question – ‘do plantsbenefit from their induced defensiveresponses?’ – has proved the most dif-ficult to answer and only a few welldesigned field studies have been performed (Karban et al., 1999).Agrawal (1998) estimated lifetime fitness of wild radish plants(Raphanus sativus) (as number of seeds produced multiplied by seedmass) assigned to one of three treatments: grazed plants (subjectto grazing by the caterpillar of Pieris rapae), leaf damage controls(equivalent amount of biomass removed using scissors) andoverall controls (undamaged). Damage-induced responses, bothchemical and physical, included increased concentrations ofdefensive glucosinolates and increased densities of trichomes(hair-like structures). Earwigs (Forficula spp.) and other chewingherbivores caused 100% more leaf damage on the control andartificially leaf-clipped plants than on grazed plants and there were30% more sucking green peach aphids (Myzus persicae) on the con-trol and leaf-clipped plants (Figure 9.2a, b). Induction of resistance,caused by grazing by the P. rapae caterpillars, significantly increasedthe lifetime index of fitness by more than 60% compared to thecontrol. However, leaf damage control plants (scissors) had 38%lower fitness than the overall controls, indicating the negative effectof tissue loss without the benefits of induction (Figure 9.2c).

This fitness benefit to wild radish occurred only in environ-ments containing herbivores; in their absence, an induced defens-ive response was inappropriate and the plants suffered reducedfitness (Karban et al., 1999). A similar fitness benefit has been shownin a field experiment involving wild tobacco (Nicotiana attenuata)(Baldwin, 1998). A specialist consumer of wild tobacco, the catter-pillar of Manduca sexta, is remarkable in that it not only inducesan accumulation of secondary metabolites and proteinase inhibitorswhen it feeds on wild tobacco, but it also induces the plants to

••••

plants make

defensive

responses . . .

. . . or do they?

are herbivores

really adversely

affected? . . .

. . . and do plants

really benefit?

EIPC09 10/24/05 2:01 PM Page 268


release volatile organic compounds that attract the generalistpredatory bug Geocoris pallens, which feeds on the slow movingcaterpillars (Kessler & Baldwin, 2004). Using molecular tech-niques, Zavala et al. (2004) were able to show that in the absenceof herbivory, plant genotypes that produced little or no proteinaseinhibitor grew faster and taller and produced more seed capsulesthan inhibitor-producing genotypes. Moreover, naturally occur-ring genotypes from Arizona that lacked the ability to produceproteinase inhibitors were damaged more, and sustained greaterManduca growth, in a laboratory experiment, compared withUtah inhibitor-producing genotypes (Glawe et al., 2003).

It is clear from the wild radish and wild tobacco examples thatthe evolution of inducible (plastic) responses involves significantcosts to the plant. We may expect inducible responses to be favoredby selection only when past herbivory is a reliable predictor offuture risk of herbivory and if the likelihood of herbivory is notconstant (constant herbivory should select for a fixed defensive

••••

Con

sum

ptio

n (g

; wet

mas

s)

00

0.2

0.1

Ungrazedcontrol plants

(b)

Previouslygrazed plants

P = 0.02

Phl

orot

anni

n co

nten

t (%

of d

ry m

ass)

Con

trol

00

8

6

4

2

Mom

enta

rycl

ippi

ng

Con

tinuo

uscl

ippi

ng

Litto

rina

obtu

sata

Idot

eagr

anul

osa

(a)

aa

a

b

a

Figure 9.1 (a) Phlorotannin content of Ascophyllum nodosumplants after exposure to simulated herbivory (removing tissue witha hole punch) or grazing by real herbivores of two species. Meansand standard errors are shown. Only the snail Littorina obtusatahad the effect of inducing increased concentrations of thedefensive chemical in the seaweed. Different letters indicate thatmeans are significantly different (P < 0.05). (b) In a subsequentexperiment, the snails were presented with algal shoots from the control and snail-grazed treatments in (a); the snails atesignificantly less of plants with a high phlorotannin content. (After Pavia & Toth 2000.)

Leaf

are

a da

mag

ed (

%)

Apr 60

5

10

15

Apr 20

(a)

Num

ber

of a

phid

s pe

r pl

ant

Apr 60

10

30

Apr 20

(b)

Pla

nt fi

tnes

s(s

eeds

× s

eed

mas

s)

Treatment0

1

2

3

(c)

20

40

Control

DamagecontrolInduced

Sampling date

Figure 9.2 (a) Percentage of leaf area consumed by chewingherbivores and (b) number of aphids per plant, measured on two dates (April 6 and April 20) in three field treatments: overallcontrol, damage control (tissue removed by scissors) and induced(caused by grazing of caterpillars of Pieris rapae). (c) The fitness of plants in the three treatments calculated by multiplying thenumber of seeds produced by the mean seed mass (in mg). (After Agrawal, 1998.)

EIPC09 10/24/05 2:01 PM Page 269

270 CHAPTER 9

phenotype that is best for that set of conditions) (Karban et al.,1999). Of course, it is not only the costs of inducible defenses thatcan be set against fitness benefits. Constitutive defenses, such asspines, trichomes or defensive chemicals (particularly in the fam-ilies Solanaceae and Brassicaceae), also have costs that have beenmeasured (in phenotypes or genotypes lacking the defense) in termsof reductions in growth or the production of flowers, fruits orseeds (see review by Strauss et al., 2002).

9.2.3 Herbivory, defoliation and plant growth

Despite a plethora of defensive struc-tures and chemicals, herbivores stilleat plants. Herbivory can stop plant

growth, it can have a negligible effect on growth rate, and it can

do just about anything in between. Plant compensation may bea general response to herbivory or may be specific to particularherbivores. Gavloski and Lamb (2000b) tested these alternativehypotheses by measuring the biomass of two cruciferous plantsBrassica napus and Sinapis alba in response to 0, 25 and 75% defoliation of seedling plants by three herbivore species with biting and chewing mouthparts – adult flea beetles Phyllotreta cruciferae and larvae of the moths Plutella xylostella and Mamestraconfigurata. Not surprisingly, both plant species compensatedbetter for 25% than 75% defoliation. However, although defoli-ated to the same extent, both plants tended to compensate bestfor defoliation by the moth M. configurata and least for the beetleP. cruciferae (Figure 9.3). Herbivore-specific compensation mayreflect plant responses to slightly different patterns of defoliationor different chemicals in saliva that suppress growth in contrastingways (Gavloski & Lamb, 2000b).

••••

Com

pens

atio

n in

dex

–2.0

–1.5

–1.0

–0.5

0.0

0.5

B. napus: 25%

Com

pens

atio

n in

dex

–2.0

–1.5

–1.0

–0.5

0.0

0.5

B. napus: 75%*

Com

pens

atio

n in

dex

–2.0

–1.5

–1.0

–0.5

0.0

0.5

S. alba: 25%

7 14 21 28

Days after defoliation

Com

pens

atio

n in

dex

–2.0

–1.5

–1.0

–0.5

0.0

0.5

S. alba: 75%

7 14 21 28

Days after defoliation

Phyllotreta cruciferae

Plutella xylostella

Mamestra configurata

*

*

*

*

*

*Figure 9.3 Compensation of leaf biomass(mean ± SE: (loge biomass defoliated plant)– (loge of mean for control plants)) ofBrassica napus and Sinapis alba seedlingswith 25 or 75% defoliation by three species of insect (see key) in a controlledenvironment. On the vertical axis, zeroequates to perfect compensation, negativevalues to undercompensation and positivevalues to overcompensation. Meanbiomasses of defoliated plants that differsignificantly from corresponding controlsare indicated by an asterisk. (After Gavloski& Lamb, 2000b.)

timing of herbivory

is crucial

EIPC09 10/24/05 2:01 PM Page 270


In the example above, compensation, which was generally complete by 21 days after defoliation, was associated with changesin root biomass consistent with the maintenance of a constantshoot : root ratio. Many plants compensate for herbivory in thisway by altering the distribution of photosynthate in different partsof the plant. Thus, for example, Kosola et al. (2002) found thatthe concentration of soluble sugars in the young (white) fine rootsof poplars (Populus canadensis) defoliated by gypsy moth caterpil-lars (Lymantria dispar) was much lower than in undefoliatedtrees. Older roots (>1 month in age), on the other hand, showedno significant effect of defoliation.

Often, there is considerable difficulty in assessing the realextent of defoliation, refoliation and hence net growth. Close monitoring of waterlily leaf beetles (Pyrrhalta nymphaeae) grazingon waterlilies (Nuphar luteum) revealed that leaves were rapidlyremoved, but that new leaves were also rapidly produced. Morethan 90% of marked leaves on grazed plants had disappeared within17 days, while marked leaves on ungrazed plants were still com-pletely intact (Figure 9.4). However, simple counts of leaves ongrazed and ungrazed plants only indicated a 13% loss of leavesto the beetles, because of new leaf production on grazed plants.

The plants that seem most tolerantof grazing, especially vertebrate grazing,are the grasses. In most species, themeristem is almost at ground levelamongst the basal leaf sheaths, and

this major point of growth (and regrowth) is therefore usually protected from grazers’ bites. Following defoliation, new leavesare produced using either stored carbohydrates or the photosyn-thate of surviving leaves, and new tillers are also often produced.

Grasses do not benefit directly from their grazers’ attentions.But it is likely that they are helped by grazers in their competit-ive interactions with other plants (which are more stronglyaffected by the grazers), accounting for the predominance ofgrasses in so many natural habitats that suffer intense vertebrategrazing. This is an example of the most widespread reason forherbivory having a more drastic effect on grazing-intolerantspecies than is initially apparent – the interaction between herbivory and plant competition (the range of possible con-sequences of which are discussed by Pacala & Crawley, 1992; see also Hendon & Briske, 2002). Note also that herbivores canhave severe nonconsumptive effects on plants when they act as vectors for plant pathogens (bacteria, fungi and especiallyviruses) – what the herbivores take from the plant is far less import-ant than what they give it! For instance, scolytid beetles feedingon the growing twigs of elm trees act as vectors for the fungusthat causes Dutch elm disease. This killed vast numbers of elmsin northeastern USA in the 1960s, and virtually eradicated themin southern England in the 1970s and early 1980s.

9.2.4 Herbivory and plant survival

Generally, it is more usual for herbivoresto increase a plant’s susceptibility tomortality than to kill it outright. Forexample, although the flea beetleAltica sublicata reduced the growth rate of the sand-dune willowSalix cordata in both 1990 and 1991 (Figure 9.5), significant mortality as a result of drought stress only occurred in 1991. Then, however, susceptibility was strongly influenced by theherbivore: 80% of plants died in a high herbivory treatment(eight beetles per plant), 40% died at four beetles per plant, butnone of the beetle-free control plants died (Bach, 1994).

Repeated defoliation can have anespecially drastic effect. Thus, a singledefoliation of oak trees by the gypsymoth (Lymantria dispar) led to only a 5%mortality rate whereas three succes-sive heavy defoliations led to mortality rates of up to 80%(Stephens, 1971). The mortality of established plants, however,is not necessarily associated with massive amounts of defoliation.One of the most extreme cases where the removal of a smallamount of plant has a disproportionately profound effect is ring-barking of trees, for example by squirrels or porcupines. Thecambial tissues and the phloem are torn away from the woodyxylem, and the carbohydrate supply link between the leaves and the roots is broken. Thus, these pests of forestry plantationsoften kill young trees whilst removing very little tissue. Surface-feeding slugs can also do more damage to newly established grass populations than might be expected from the quantity ofmaterial they consume (Harper, 1977). The slugs chew through

••••

Ungrazed Grazed

170

1

80

100

11(Jul 26)

4(Aug 11)

Days since marking

60

40

20

Leaf

are

a re

mai

ning

(%

)

Figure 9.4 The survivorship of leaves on waterlily plants grazedby the waterlily leaf beetle was much lower than that on ungrazedplants. Effectively, all leaves had disappeared at the end of 17 days,despite the fact that ‘snapshot’ estimates of loss rates to grazing ongrazed plants during this period suggested only around a 13% loss.(After Wallace & O’Hop, 1985.)

grasses are

particularly tolerant

of grazing

mortality: the result

of an interaction with

another factor?

repeated defoliation

or ring-barking

can kill

EIPC09 10/24/05 2:01 PM Page 271

272 CHAPTER 9

the young shoots at ground level, leaving the felled leavesuneaten on the soil surface but consuming the meristematicregion at the base of shoots from which regrowth would occur.They therefore effectively destroy the plant.

Predation of seeds, not surprisingly, has a predictably harmful effect on individual plants (i.e. the seeds themselves).Davidson et al. (1985) demonstrated dramatic impacts of seed-eating ants and rodents on the composition of seed banks of ‘annual’plants in the deserts of southwestern USA and thus on the makeup of the plant community.

9.2.5 Herbivory and plant fecundity

The effects of herbivory on plantfecundity are, to a considerable extent,reflections of the effects on plantgrowth: smaller plants bear fewer seeds.However, even when growth appearsto be fully compensated, seed produc-

tion may nevertheless be reduced because of a shift of resourcesfrom reproductive output to shoots and roots. This was the case in the study shown in Figure 9.3 where compensation ingrowth was complete after 21 days but seed production was stillsignificantly lower in the herbivore-damaged plants. Moreover,indirectly through its effect on leaf area, or by directly feedingon reproductive structures, herbivory can affect floral traits(corolla diameter, floral tube length, flower number) and havean adverse impact on pollination and seed set (Mothershead &

Marquis, 2000). Thus experimentally ‘grazed’ plants of Oenotheramacrocarpa produced 30% fewer flowers and 33% fewer seeds.

Plants may also be affected moredirectly, by the removal or destructionof flowers, flower buds or seeds. Thus,caterpillars of the large blue butterflyMaculinea rebeli feed only in the flowersand on the fruits of the rare plantGentiana cruciata, and the number of seeds per fruit (70 comparedto 120) is reduced where this specialist herbivore occurs (Kery et al., 2001). Many studies, involving the artificial exclusion orremoval of seed predators, have shown a strong influence of predispersal seed predation on recruitment and the density of attacked species. For example, seed predation was a significantfactor in the pattern of increasing abundance of the shrubHaplopappus squarrosus along an elevational gradient from theCalifornian coast, where predispersal seed predation was higher,to the mountains (Louda, 1982); and restriction of the cruciferCardamine cordifolia to shaded situations in the Rocky Mountainswas largely due to much higher levels of predispersal seed pre-dation in unshaded locations (Louda & Rodman, 1996).

It is important to realize, however,that many cases of ‘herbivory’ of reprod-uctive tissues are actually mutualistic,benefitting both the herbivore and theplant (see Chapter 13). Animals that‘consume’ pollen and nectar usually transfer pollen inadvertentlyfrom plant to plant in the process; and there are many fruit-eating animals that also confer a net benefit on both the parent

••••

No herbivory

Low herbivory

High herbivory

Clone number

41

0.8

32

Rel

ativ

e ch

ange

in h

eigh

t

0.6

0.4

0.2

0.05 6

0.6

87

0.4

0.2

0.09

(b) Aug 10 – Aug 21(a) Jul 19 – Aug 17

Figure 9.5 Relative growth rates (changes in height, with standard errors) of a number of different clones of the sand-dune willow, Salix cordata, (a) in 1990 and (b) in 1991, subjected either to no herbivory, low herbivory (four flea beetles per plant) or high herbivory(eight beetles per plant). (After Bach, 1994.)

herbivores affect

plant growth . . .

. . . indirectly by

reducing seed

production . . .

. . . and directly

by removing

reproductive

structures

much pollen and

fruit herbivory

benefits the plant

EIPC09 10/24/05 2:01 PM Page 272


plant and the individual seed within the fruit. Most vertebrate fruit-eaters, in particular, either eat the fruit but discard the seed, oreat the fruit but expel the seed in the feces. This disperses the seed,rarely harms it and frequently enhances its ability to germinate.

Insects that attack fruit or developing fruit, on the otherhand, are very unlikely to have a beneficial effect on the plant.They do nothing to enhance dispersal, and they may even makethe fruit less palatable to vertebrates. However, some large ani-mals that normally kill seeds can also play a part in dispersing them,and they may therefore have at least a partially beneficial effect.There are some ‘scatter-hoarding’ species, like certain squirrels,that take nuts and bury them at scattered locations; and there areother ‘seed-caching’ species, like some mice and voles, that collectscattered seeds into a number of hidden caches. In both cases,although many seeds are eaten, the seeds are dispersed, they arehidden from other seed predators and a number are never relocated by the hoarder or cacher (Crawley, 1983).

Herbivores also influence fecundity in a number of otherways. One of the most common responses to herbivore attack isa delay in flowering. For instance, in longer lived semelparousspecies, herbivory frequently delays flowering for 1 year ormore, and this typically increases the longevity of such plants since

death almost invariably follows their single burst of reproduction(see Chapter 4). Poa annua on a lawn can be made almostimmortal by mowing it at weekly intervals, whereas in naturalhabitats, where it is allowed to flower, it is commonly an annual– as its name implies.

Generally, the timing of defoliationis critical in determining the effect onplant fecundity. If leaves are removedbefore inflorescences are formed, then the extent to whichfecundity is depressed clearly depends on the extent to which theplant is able to compensate. Early defoliation of a plant with sequen-tial leaf production may have a negligible effect on fecundity; but where defoliation takes place later, or where leaf productionis synchronous, flowering may be reduced or even inhibitedcompletely. If leaves are removed after the inflorescence hasbeen formed, the effect is usually to increase seed abortion or toreduce the size of individual seeds.

An example where timing is important is provided by field gen-tians (Gentianella campestris). When herbivory on this biennial plantis simulated by clipping to remove half its biomass (Figure 9.6a),the outcome depends on the timing of the clipping (Figure 9.6b).Fruit production was much increased over controls if clipping

••••

Unclipped Clipped

Before clipping

(a)

(b)

Jul 1

20

Contro

l

30

Num

ber

of fr

uits

25

20

15

10

5

Jul 2

0

Jul 2

8

a

b

c

d

the timing of

herbivory is critical

Figure 9.6 (a) Clipping of field gentiansto simulate herbivory causes changes in the architecture and numbers of flowersproduced. (b) Production of mature (openhistograms) and immature fruits (blackhistograms) of unclipped control plants andplants clipped on different occasions fromJuly 12 to 28, 1992. Means and standarderrors are shown and all means aresignificantly different from each other (P < 0.05). Plants clipped on July 12 and 20 developed significantly more fruits thanunclipped controls. Plants clipped on July28 developed significantly fewer fruits thancontrols. (After Lennartsson et al., 1998).

EIPC09 10/24/05 2:01 PM Page 273

274 CHAPTER 9

occurred between 1 and 20 July, but if clipping occurred later thanthis, fruit production was less in the clipped plants than in theunclipped controls. The period when the plants show compen-sation coincides with the time when damage by herbivores nor-mally occurs.

9.2.6 A postscript: antipredator chemical defenses in animals

It should not be imagined that antipred-ator chemical defenses are restricted toplants. A variety of constitutive animal

chemical defenses were described in Chapter 3 (see Section 3.7.4),including plant defensive chemicals sequestered by herbivores fromtheir food plants (see Section 3.7.4). Chemical defenses may be particularly important in modular animals, such as sponges,which lack the ability to escape from their predators. Despite theirhigh nutritional value and lack of physical defenses, most marinesponges appear to be little affected by predators (Kubanek et al.,2002). In recent years, several triterpene glycosides have beenextracted from sponges, including from Ectyoplasia ferox in theCaribbean. In a field study, crude extracts of refined triterpeneglycosides from this sponge were presented in artificial food substrates to natural assemblages of reef fishes in the Bahamas.Strong antipredatory affects were detected when compared to control substrates (Figure 9.7). It is of interest that the triterpeneglycosides also adversely affected competitors of the sponge, includ-ing ‘fouling’ organisms that overgrow them (bacteria, invertebratesand algae) and other sponges (an example of allelopathy – seeSection 8.3.2). All these enemies were apparently deterred by surface contact with the chemicals rather than by water-borneeffects (Kubanek et al., 2002).

9.3 The effect of predation on prey populations

Returning now to predators in general, it may seem that since the effects of predators are harmful to individual prey, theimmediate effect of predation on a population of prey must alsobe predictably harmful. However, these effects are not always sopredictable, for one or both of two important reasons. In the firstplace, the individuals that are killed (or harmed) are not alwaysa random sample of the population as a whole, and may be thosewith the lowest potential to contribute to the population’s future.Second, there may be compensatory changes in the growth, sur-vival or reproduction of the surviving prey: they may experiencereduced competition for a limiting resource, or produce more off-spring, or other predators may take fewer of the prey. In otherwords, whilst predation is bad for the prey that get eaten, it maybe good for those that do not. Moreover, predation is least likelyto affect prey dynamics if it occurs at a stage of the prey’s lifecycle that does not have a significant effect, ultimately, on preyabundance.

To deal with the second point first,if, for example, plant recruitment isnot limited by the number of seedsproduced, then insects that reduceseed production are unlikely to have an important effect on plant abundance (Crawley, 1989). For instance, the weevilRhinocyllus conicus does not reduce recruitment of the noddingthistle, Carduus nutans, in southern France despite inflicting seed losses of over 90%. Indeed, sowing 1000 thistle seeds per square meter also led to no observable increase in the numberof thistle rosettes. Hence, recruitment appears not to be limitedby the number of seeds produced; although whether it is limited by subsequent predation of seeds or early seedlings, orthe availability of germination sites, is not clear (Crawley, 1989).(However, we have seen in other situations (see Section 9.2.5)that predispersal seed predation can profoundly affect seed-ling recruitment, local population dynamics and variation in relative abundance along environmental gradients and acrossmicrohabitats.)

The impact of predation is oftenlimited by compensatory reactionsamongst the survivors as a result ofreduced intraspecific competition. Thus,in a classic experiment in which large numbers of woodpigeons(Columba palumbus) were shot, the overall level of winter mor-tality was not increased, and stopping the shooting led to noincrease in pigeon abundance (Murton et al., 1974). This wasbecause the number of surviving pigeons was determined ultimatelynot by shooting but by food availability, and so when shootingreduced density, there were compensatory reductions in intra-specific competition and in natural mortality, as well as density-dependent immigration of birds moving in to take advantage ofunexploited food.

••••

% e

aten

0

100

Control

(a)

Treated

80

60

40

20

0

100

Control

(b)

Treated

80

60

40

20

Figure 9.7 Results of field assays assessing antipredatory effectsof compounds from the sponge Ectyoplasia ferox against naturalassemblages of reef fish in the Bahamas. Means (+ SE) are shownfor percentages of artificial food substrates eaten in controls(containing no sponge extracts) in comparison with: (a) substratescontaining a crude sponge extract (t-test, P = 0.036) and (b) substrates containing triterpene glycosides from the sponge (P = 0.011). (After Kubanek et al., 2002.)

animals also defend

themselves

predation may occur

at a demographically

unimportant stage

compensatory

reactions amongst

survivors

EIPC09 10/24/05 2:01 PM Page 274


Indeed, whenever density is highenough for intraspecific competitionto occur, the effects of predation on apopulation should be ameliorated by the

consequent reductions in intraspecific competition. Outcomes ofpredation may, therefore, vary with relative food availability. Wherefood quantity or quality is higher, a given level of predation maynot lead to a compensatory response because prey are not food-limited. This hypothesis was tested by Oedekoven and Joern(2000) who monitored grasshopper (Ageneotettix deorum) sur-vivorship in caged prairie plots subject to fertilization (or not) to increase food quality in the presence or absence of lycosid spiders (Schizocoza spp.). With ambient food quality (no fertilizer,black symbols), spider predation and food limitation were com-pensatory: the same numbers of grasshoppers were recovered at the end of the 31-day experiment (Figure 9.8). However, withhigher food quality (nitrogen fertilizer added, colored symbols), spider predation reduced the numbers surviving compared to theno-spider control: a noncompensatory response. Under ambientconditions after spider predation, the surviving grasshoppersencountered more food per capita and lived longer as a result ofreduced competition. However, grasshoppers were less food-limited when food quality was higher so that after predation therelease of additional per capita food did not promote survivor-ship (Oedekoven & Joern, 2000).

Turning to the nonrandom distribu-tion of predators’ attention within a population of prey, it is likely, forexample, that predation by many largecarnivores is focused on the old (and

infirm), the young (and naive) or the sick. For instance, a study

in the Serengeti found that cheetahs and wild dogs killed a dispro-portionate number from the younger age classes of Thomson’sgazelles (Figure 9.9a), because: (i) these young animals were easier to catch (Figure 9.9b); (ii) they had lower stamina and running speeds; (iii) they were less good at outmaneuvering the predators (Figure 9.9c); and (iv) they may even have failed to recognize the predators (FitzGibbon & Fanshawe, 1989;FitzGibbon, 1990). Yet these young gazelles will also have beenmaking no reproductive contribution to the population, and theeffects of this level of predation on the prey population willtherefore have been less than would otherwise have been the case.

Similar patterns may also be found in plant populations. Themortality of mature Eucalyptus trees in Australia, resulting fromdefoliation by the sawfly Paropsis atomaria, was restricted almostentirely to weakened trees on poor sites, or to trees that had suffered from root damage or from altered drainage following cultivation (Carne, 1969).

Taken overall, then, it is clear thatthe step from noting that individualprey are harmed by individual predatorsto demonstrating that prey adundanceis adversely affected is not an easy one to take. Of 28 studies inwhich herbivorous insects were experimentally excluded from plantcommunities using insecticides, 50% provided evidence of an effecton plants at the population level (Crawley, 1989). As Crawley noted,however, such proportions need to be treated cautiously. There isan almost inevitable tendency for ‘negative’ results (no popula-tion effect) to go unreported, on the grounds of there being ‘nothing’ to report. Moreover, the exclusion studies often took 7 years or more to show any impact on the plants: it may be that many of the ‘negative’ studies were simply given up too early.

••••

No spiders, no fertilizer

No spiders, fertilizer

Spiders, no fertilizer

Spiders, fertilizer

Log e

(num

ber

of g

rass

hopp

ers)

201550

0

1

2

3

10

Time (days)

25 30 35

Figure 9.8 Trajectories of numbers of grasshoppers surviving (mean ± SE) for fertilizer and predation treatmentcombinations in a field experimentinvolving caged plots in the ArapahoPrairie, Nebraska, USA. (After Oedekoven & Joern, 2000.)

effects ameliorated

by reduced

competition

predatory attacks are

often directed at the

weakest prey

difficulties of

demonstrating effects

on prey populations

EIPC09 10/24/05 2:01 PM Page 275

••

276 CHAPTER 9

Many more recent investigations have shown clear effects of seed predation on plant abundance (e.g. Kelly & Dyer, 2002; Maronet al., 2002).

9.4 Effects of consumption on consumers

The beneficial effects that food has onindividual predators are not difficult to imagine. Generally speaking, anincrease in the amount of food con-sumed leads to increased rates of

growth, development and birth, and decreased rates of mortal-ity. This, after all, is implicit in any discussion of intraspecific competition amongst consumers (see Chapter 5): high densities,implying small amounts of food per individual, lead to lowgrowth rates, high death rates, and so on. Similarly, many of theeffects of migration previously considered (see Chapter 6) reflectthe responses of individual consumers to the distribution of foodavailability. However, there are a number of ways in which therelationships between consumption rate and consumer benefit can be more complicated than they initially appear. In the firstplace, all animals require a certain amount of food simply for maintenance and unless this threshold is exceeded the animal will be unable to grow or reproduce, and will therefore beunable to contribute to future generations. In other words, lowconsumption rates, rather than leading to a small benefit to theconsumer, simply alter the rate at which the consumer starvesto death.

At the other extreme, the birth,growth and survival rates of individualconsumers cannot be expected to riseindefinitely as food availability is increased. Rather, the con-sumers become satiated. Consumption rate eventually reaches aplateau, where it becomes independent of the amount of food avail-able, and benefit to consumers therefore also reaches a plateau.Thus, there is a limit to the amount that a particular consumerpopulation can eat, a limit to the amount of harm that it can do to its prey population at that time, and a limit to the extentby which the consumer population can increase in size. This isdiscussed more fully in Section 10.4.

The most striking example of wholepopulations of consumers being sati-ated simultaneously is provided by the many plant species that have mastyears. These are occasional years in which there is synchronousproduction of a large volume of seed, often across a large geo-graphic area, with a dearth of seeds produced in the years inbetween (Herrera et al., 1998; Koenig & Knops, 1998; Kelly et al.,2000). This is seen particularly often in tree species that suffer gen-erally high intensities of seed predation (Silvertown, 1980) and itis therefore especially significant that the chances of escaping seedpredation are likely to be much higher in mast years than in otheryears. Masting seems to be especially common in the NewZealand flora (Kelly, 1994) where it has also been reported fortussock grass species (Figure 9.10). The individual predators of seedsare satiated in mast years, and the populations of predators can-not increase in abundance rapidly enough to exploit the glut. This

••

Per

cent

age

0

Faw

ns

40

60

80

(a)

20

Hal

f-gro

wns

Adol

esce

nts

Sub-

adul

ts

Adul

ts

Killed by cheetahs

Killed by wild dogs

Percentage in population

Per

cent

age

of c

hase

dga

zelle

s es

capi

ng

0

Faw

ns

40

60

80

(b)

20

Hal

f-gro

wns

Adol

esce

nts

Dis

tanc

e lo

st (

m)

–1.5

Faw

ns

0.0

1.0

2.0

(c)

–1.0

Hal

f-gro

wns

and

adol

esce

nts

Adul

ts

–0.5

0.5

1.5

Figure 9.9 (a) The proportions of different age classes (determined by tooth wear) of Thomson’s gazelles in cheetah and wild dog kills isquite different from their proportions in the population as a whole. (b) Age influences the probability for Thomson’s gazelles of escapingwhen chased by cheetahs. (c) When prey (Thomson’s gazelles) ‘zigzag’ to escape chasing cheetahs, prey age influences the mean distancelost by the cheetahs. (After FitzGibbon & Fanshawe, 1989; FitzGibbon, 1990.)

consumers often

need to exceed

a threshold of

consumption

consumers may

become satiated

mast years and the

satiation of seed

predators

EIPC09 10/24/05 2:01 PM Page 276

••


is illustrated in Figure 9.11 where the percentage of florets of thegrass Chionochloa pallens attacked by insects remains below 20%in mast years but ranges up to 80% or more in nonmast years.The fact that C. pallens and four other species of Chionochloa showstrong synchrony in masting is likely to result in an increased benefitto each species in terms of escaping seed predation in mast years.

On the other hand, the production of a mast crop makes greatdemands on the internal resources of a plant. A spruce tree in amast year averages 38% less annual growth than in other years,and the annual ring increment in forest trees may be reduced byas much during a mast year as by a heavy attack of defoliatingcaterpillars. The years of seed famine are therefore essentially yearsof plant recovery.

As well as illustrating the potentialimportance of predator satiation, theexample of masting highlights a furtherpoint relating to timescales. The seedpredators are unable to extract themaximum benefit from (or do the maximum harm to) the mastcrop because their generation times are too long. A hypotheticalseed predator population that could pass through several gener-ations during a season would be able to increase exponentiallyand explosively on the mast crop and destroy it. Generally speak-ing, consumers with relatively short generation times tend to closelytrack fluctuations in the quantity or abundance of their food or

••

Flo

wer

ing

inte

nsity

(inflo

resc

ence

s tu

ssoc

k–1 )

199519850

1975

10

20

30

1980

5

15

25

1990

C. rubraC. seretofoliaC. rigida

Flo

wer

ing

inte

nsity

(inflo

resc

ence

s tu

ssoc

k–1 )

199519850

1975

4

6

8

1980

Year

1990

2

C. crassiusculaC. palliens

Mast years0

20

Nonmast years

40

60

80

Inse

ct p

reda

tion

(% fl

oret

s at

tack

ed)

Figure 9.10 The flowering rate for fivespecies of tussock grass (Chionochloa)between 1973 and 1996 in FiordlandNational Park, New Zealand. Mast yearsare highly synchronized in the five species,seemingly in response to high temperaturesin the previous season, when flowering isinduced. (After McKone et al., 1998.)

Figure 9.11 Insect predation on florets of Chionochloa pallensin mast (n = 3) and nonmast years (n = 7) from 1988 to 1997 atMount Hutt, New Zealand. A mast year is defined here as onewith greater than 10 times as many florets produced per tussockthan in the previous year. The significant difference in insectdamage supports the hypothesis that the function of masting is to satiate seed predators. (After McKone et al., 1998.)

a consumer’s

numerical response

is limited by its

generation time . . .

EIPC09 10/24/05 2:01 PM Page 277

278 CHAPTER 9

prey, whereas consumers with relatively long generation timestake longer to respond to increases in prey abundance, andlonger to recover when reduced to low densities.

The same phenomenon occurs indesert communities, where year-to-year variations in precipitation can beboth considerable and unpredictable,

leading to similar year-to-year variation in the productivity of manydesert plants. In the rare years of high productivity, herbivoresare typically at low abundance following one or more years oflow plant productivity. Thus, the herbivores are likely to be sati-ated in such years, allowing plant populations to add consider-ably to their reserves, perhaps by augmenting their buried seedbanks or their underground storage organs (Ayal, 1994). The ex-ample of fruit production by Asphodelus ramosus in the Negev desertin Israel in shown in Figure 9.12. The mirid bug, Capsodes infus-catus, feeds on Asphodelus, exhibiting a particular preference forthe developing flowers and young fruits. Potentially, therefore,it can have a profoundly harmful effect on the plant’s fruit production. But it only passes through one generation per year.Hence, its abundance tends never to match that of its host plant(Figure 9.12). In 1988 and 1991, fruit production was high but mirid abundance was relatively low: the reproductive output of the mirids was therefore high (3.7 and 3.5 nymphs per adult,respectively), but the proportion of fruits damaged was relativelylow (0.78 and 0.66). In 1989 and 1992, on the other hand, whenfruit production had dropped to much lower levels, the propor-tion of fruits damaged was much higher (0.98 and 0.87) and thereproductive output was lower (0.30 nymphs per adult in 1989;unknown in 1992). This suggests that herbivorous insects, at least,may have a limited ability to affect plant population dynamics in desert communities, but that the potential is much greater forthe dynamics of herbivorous insects to be affected by their foodplants (Ayal, 1994).

Chapter 3 stressed that the quantityof food consumed may be less import-ant than its quality. In fact, food qual-ity, which has both positive aspects(like the concentrations of nutrients)and negative aspects (like the concentrations of toxins), can onlysensibly be defined in terms of the effects of the food on the animal that eats it; and this is particularly pertinent in the case of herbivores. For instance, we saw in Figure 9.8 how even inthe presence of predatory spiders, enhanced food quality led toincreased survivorship of grasshoppers. Along similar lines,Sinclair (1975) examined the effects of grass quality (protein con-tent) on the survival of wildebeest in the Serengeti of Tanzania.Despite selecting protein-rich plant material (Figure 9.13a), thewildebeest consumed food in the dry season that contained wellbelow the level of protein necessary even for maintenance (5–6%of crude protein); and to judge by the depleted fat reserves of deadmales (Figure 9.13b), this was an important cause of mortality.Moreover, it is highly relevant that the protein requirements offemales during late pregnancy and lactation (December–May inthe wildebeest) are three to four times higher than the normal.It is therefore clear that the shortage of high-quality food (andnot just food shortage per se) can have a drastic effect on the growth,survival and fecundity of a consumer. In the case of herbivoresespecially, it is possible for an animal to be apparently surroundedby its food whilst still experiencing a food shortage. We can seethe problem if we imagine that we ourselves are provided witha perfectly balanced diet – diluted in an enormous swimming pool.The pool contains everything we need, and we can see it therebefore us, but we may very well starve to death before we candrink enough water to extract enough nutrients to sustain our-selves. In a similar fashion, herbivores may frequently be confrontedwith a pool of available nitrogen that is so dilute that they havedifficulty processing enough material to extract what they need.Outbreaks of herbivorous insects may then be associated with rareelevations in the concentration of available nitrogen in their foodplants (see Section 3.7.1), perhaps associated with unusually dryor, conversely, unusually waterlogged conditions (White, 1993).Consumers obviously need to acquire resources – but, to benefitfrom them fully they need to acquire them in appropriate quant-ities and in an appropriate form. The behavioral strategies thathave evolved in the face of the pressures to do this are the maintopic of the next two sections.

9.5 Widths and compositions of diets

Consumers can be classified as eithermonophagous (feeding on a singleprey type), oligophagous (few preytypes) or polyphagous (many preytypes). An equally useful distinction is

••••

Num

ber

of in

divi

dual

s (1

000s

)

939290087

2.1

2.8

3.5

91

Year

1.4

0.7

88 89

Num

ber

of fr

uits

(10

00s)

0

30

20

10

Figure 9.12 Fluctuations in the fruit production of Asphodelus (�)and the number of Capsodes nymphs (�) and adults (�) at a studysite in the Negev desert, Israel. (After Ayal, 1994.)

. . . as illustrated by

desert interactions

food quality rather

than quantity can

be of paramount

importance

range and

classification of

diet widths

EIPC09 10/24/05 2:01 PM Page 278


between specialists (broadly, monophages and oligophages) andgeneralists (polyphages). Herbivores, parasitoids and true preda-tors can all provide examples of monophagous, oligophagous andpolyphagous species. But the distribution of diet widths differsamongst the various types of consumer. True predators with spe-cialized diets do exist (for instance the snail kite Rostrahamus socia-bilis feeds almost entirely on snails of the genus Pomacea), but mosttrue predators have relatively broad diets. Parasitoids, on the otherhand, are typically specialized and may even be monophagous.Herbivores are well represented in all categories, but whilst grazing and ‘predatory’ herbivores typically have broad diets, ‘par-asitic’ herbivores are very often highly specialized. For instance,Janzen (1980) examined 110 species of beetle that feed as larvaeinside the seeds of dicotyledonous plants in Costa Rica (‘parasitizing’them) and found that 83 attacked only one plant species, 14attacked only two, nine attacked three, two attacked four, oneattacked six and one attacked eight of the 975 plants in the area.

9.5.1 Food preferences

It must not be imagined that poly-phagous and oligophagous species areindiscriminate in what they choosefrom their acceptable range. On the

contrary, some degree of preference is almost always apparent.An animal is said to exhibit a preference for a particular type offood when the proportion of that type in the animal’s diet is higherthan its proportion in the animal’s environment. To measure

food preference in nature, therefore, it is necessary not only toexamine the animal’s diet (usually by the analysis of gut contents)but also to assess the ‘availability’ of different food types. Ideally,this should be done not through the eyes of the observer (i.e. notby simply sampling the environment), but through the eyes ofthe animal itself.

A food preference can be expressed in two rather different con-texts. There can be a preference for items that are the most valu-able amongst those available or for items that provide an integralpart of a mixed and balanced diet. These will be referred to asranked and balanced preferences, respectively. In the terms ofChapter 3 (Section 3.8), where resources were classified, indi-viduals exhibit ranked preferences in discriminating between re-source types that are ‘perfectly substitutable’ and exhibit balancedpreferences between resource types that are ‘complementary’.

Ranked preferences are usuallyseen most clearly amongst carnivores.For instance, Figure 9.14 shows twoexamples in which carnivores activelyselected prey items that were the mostprofitable in terms of energy intakeper unit time spent dealing with (or‘handling’) prey. Results such as these reflect the fact that a car-nivore’s food often varies little in composition (see Section 3.7.1),but may vary in size or accessibility. This allows a single meas-ure (like ‘energy gained per unit handling time’) to be used tocharacterize food items, and it therefore allows food items to beranked. In other words, Figure 9.14 shows consumers exhibitingan active preference for food of a high rank.

••••

Cru

de p

rote

in (

%)

0N

5

10

20(a)

15

D J F M A M J J A S O

Bon

e m

arro

w fa

t (%

)

0N

50

100(b)

D J A M J J A S OF M

Figure 9.13 (a) The quality of food measured as percentage crude protein available to (7) and eaten by (�) wildebeest in the Serengetiduring 1971. Despite selection (‘eaten’ > ‘available’), the quality of food eaten fell during the dry season below the level necessary for themaintenance of nitrogen balance (5–6% of crude protein). (b) The fat content of the bone marrow of the live male population (7) andthose found dead from natural causes (�). Vertical lines, where present, show 95% confidence limits. (After Sinclair, 1975.)

preference is defined

by comparing diet

with ‘availability’

ranked preferences

predominate when

food items can be

classified on a single

scale . . .

EIPC09 10/24/05 2:01 PM Page 279

280 CHAPTER 9

For many consumers, however,especially herbivores and omnivores,no simple ranking is appropriate, sincenone of the available food itemsmatches the nutritional requirements of the consumer. These requirements

can therefore only be satisfied either by eating large quantities of food, and eliminating much of it in order to get a sufficientquantity of the nutrient in most limited supply (for exampleaphids and scale insects excrete vast amounts of carbon in honeydew to get sufficient nitrogen from plant sap), or by eatinga combination of food items that between them match the con-sumer’s requirements. In fact, many animals exhibit both sortsof response. They select food that is of generally high quality (so the proportion eliminated is minimized), but they also selectitems to meet specific requirements. For instance, sheep and cattle show a preference for high-quality food, selecting leaves in preference to stems, green matter in preference to dry or old material, and generally selecting material that is higher in nitrogen, phosphorus, sugars and gross energy, and lower infiber, than what is generally available. In fact, all generalist herbivores appear to show rankings in the rate at which they eatdifferent food plants when given a free choice in experimental tests(Crawley, 1983).

On the other hand, a balanced preference is also quite common. Forinstance, the plate limpet, Acmaea scutum, selects a diet of two species of encrusting microalgae that contains

60% of one species and 40% of the other, almost irrespective ofthe proportions in which they are available (Kitting, 1980). Whilstcaribou, which survive on lichen through the winter, develop a

sodium deficiency by the spring that they overcome by drinkingseawater, eating urine-contaminated snow and gnawing shedantlers (Staaland et al., 1980). We have only to look at ourselvesto see an example in which ‘performance’ is far better on amixed diet than on a pure diet of even the ‘best’ food.

There are two other important reasons why a mixed diet maybe favored. First, consumers may accept low-quality items sim-ply because, having encountered them, they have more to gainby eating them (poor as they are) than by ignoring them and con-tinuing to search. This is discussed in detail in Section 9.5.3. Second,consumers may benefit from a mixed diet because each food typemay contain a different undesirable toxic chemical. A mixed dietwould then keep the concentrations of all of these chemicals withinacceptable limits. It is certainly the case that toxins can play animportant role in food preference. For instance, dry matterintake by Australian ringtail possums (Pseudocheirus peregrinus) feed-ing on Eucalyptus tree leaves was strongly negatively correlatedwith the concentration of sideroxylonal, a toxin found inEucalyptus leaves, but was not related to nutritional character-istics such as nitrogen or cellulose (Lawler et al., 2000).

Overall, however, it would be quite wrong to give theimpression that all preferences have been clearly linked with oneexplanation or another. For example, Thompson (1988) reviewedthe relationship between the oviposition preferences of phy-tophagous insects and the performance of their offspring on theselected food plants in terms of growth, survival and reproduc-tion. A number of studies have shown a good association (i.e.females preferentially oviposit on plants where their offspring perform best), but in many others the association is poor. In such cases there is generally no shortage of explanations for theapparently unsuitable behavior, but these explanations are, as yet,often just untested hypotheses.

••••

Flies selected

Flies availableE

nerg

y ga

in (

J s–

1 )

4030100

0

2.0

4.0

6.0

20

Length of mussel (mm)

(a)

Num

ber

of m

usse

lsea

ten

per

day

5

0

4321

7

Prey length (mm)

(b)

Energy value

8

Cal

orie

s s–

1 ha

ndlin

g tim

e

1096

10

5

12

14

16

Fre

quen

cy (

%)

10960

5

10

30

50

7

Prey length (mm)

40

20

8

Energy

Figure 9.14 Predators eating ‘profitable’ prey, i.e. predators showing a preponderance in their diet for those prey items that provide themwith the most energy. (a) When crabs (Carcinus maenas) were presented with equal quantities of six size classes of mussels (Mytilus edulis),they tended to show a preference for those providing the greatest energy gain (energy per unit handling time). (After Elner & Hughes,1978.) (b) Pied wagtails (Motacilla alba yarrellii) tended to select, from scatophagid flies available, those providing the greatest energy gainper unit handling time. (After Davies, 1977; Krebs, 1978.)

. . . but many

consumers show

a combination of

ranked and balanced

preferences

mixed diets can be

favored for a variety

of reasons

EIPC09 10/24/05 2:01 PM Page 280


9.5.2 Switching

The preferences of many consumersare fixed; in other words, they aremaintained irrespective of the relativeavailabilities of alternative food types.But many others switch their preference,

such that food items are eaten disproportionately often when theyare common and are disproportionately ignored when they arerare. The two types of preference are contrasted in Figure 9.15.Figure 9.15a shows the fixed preference exhibited by predatoryshore snails when they were presented with two species of mussel prey at a range of proportions. The line in Figure 9.15ahas been drawn on the assumption that they exhibited the samepreference at all proportions. This assumption is clearly justified:irrespective of availability, the predatory snails showed the samemarked preference for the thin-shelled, less protected Mytilusedulis, which they could exploit more effectively. By contrast,

Figure 9.15b shows what happened when guppies (fish) wereoffered a choice between fruit-flies and tubificid worms as prey.The guppies clearly switched their preference, and consumed adisproportionate number of the more abundant prey type.

There are a number of situations inwhich switching can arise. Probablythe most common is where differenttypes of prey are found in differentmicrohabitats, and the consumers concentrate on the mostprofitable microhabitat. This was the case for the guppies inFigure 9.15b: the fruit-flies floated at the water surface whilst thetubificids were found at the bottom. Switching can also occur(Bergelson, 1985) in the following situations:

1 When there is an increased probability of orientating towarda common prey type, i.e. consumers develop a ‘search image’for abundant food (Tinbergen, 1960) and concentrate on their‘image’ prey to the relative exclusion of nonimage prey.

••••

M. e

dulis

eat

en (

%)

10080400

0

40

80

100

60

M. edulis offered (%)

(a)

20

60

20

Expected if nopreference

Pro

port

ion

of tu

bific

ids

in d

iet

0.80.40

0

0.4

0.8

1.0

0.6

Proportion of tubificids available

(b)

0.2

0.6

0.2

Expected if nopreference

Pro

port

ion

ofG

amm

arus

eat

en1.0

00

1.0

Proportion of Gammarus available

(d)

0.5

0.5

Num

ber

of g

uppi

es

1.00.80.40

0

4

8

0.6

Proportion of tubificids in diet

(c)

2

6

0.2

Figure 9.15 Switching. (a) A lack of switching: snails exhibit a consistent preference amongst the mussels Mytilus edulis and M. californianus, irrespective of their relative abundance (means plus standard errors). (After Murdoch & Stewart-Oaten, 1975.) (b) Switching by guppies fedon tubificids and fruit-flies: they take a disproportionate amount of whichever prey type is the more available (means and total ranges).(After Murdoch et al., 1975.) (c) Preferences shown by the individual guppies in (b) when offered equal amounts of the two prey types:individuals were mostly specialists on one or other type. (d) Switching by sticklebacks fed mixtures of Gammarus and Artemia: overall theytake a disproportionate amount of whichever is more available. However, in the first series of trials, with Gammarus availability decreasing(closed symbols), first-day trialists (�) tended to take more Gammarus than third-day trialists (�), whereas with Gammarus availabilityincreasing, firsts (4) tended to take less Gammarus than thirds (7). The effects of learning are apparent. (After Hughes & Croy, 1993.)

switching involves a

preference for food

types that are

commonwhen might

switching arise?

EIPC09 10/24/05 2:01 PM Page 281

282 CHAPTER 9

2 When there is an increased probability of pursuing a commonprey type.

3 When there is an increased probability of capturing a commonprey type.

4 When there is an increased efficiency in handling a commonprey type.

In each case, increasingly common prey generate increased interest and/or success on the part of the predator, and hence anincreased rate of consumption. For instance, switching occurredin the 15-spined stickleback, Spinachia spinachia, feeding on thecrustaceans Gammarus and Artemia as alternative prey (Figure 9.15d)as a result of learned improvements in capturing and handlingefficiencies, especially of Gammarus. Fish were fed Gammarus for7 days, which was then replaced in the diet, in 10% steps, with Artemiauntil the diet was 100% Artemia. This diet was then maintainedfor a further 7 days, when the process was reversed back downto 100% Gammarus. Each ‘step’ itself lasted 3 days, on each of which the fish were tested. The learning process is apparent inFigure 9.15d in the tendency for first-day trialists to be moreinfluenced than third-day trialists by the previous dietary mix.

Interestingly, switching in a population often seems to be aconsequence not of individual consumers gradually changingtheir preference, but of the proportion of specialists changing. Figure9.15c shows this for the guppies. When the prey types were equallyabundant, individual guppies were not generalists – rather, therewere approximately equal numbers of fruit-fly and tubificid specialists.

It may come as a surprise that aplant may show behavior akin toswitching. The northern pitcher plant

Sarracenia purpurea lives in nutrient-poor bogs and fens, circum-stances that are thought to favor carnivory in plants. Carnivorousplants such as pitcher plants invest an excess of carbon (capturedin photosynthesis) in specialist organs for capturing invertebrateprey (effectively nitrogen-capturing structures). Figure 9.16 showshow relative size of the pitcher keel responded to nitrogen addi-tion to plots in Molly Bog in Vermont, USA. The more nitrogenthat was applied, the larger the relative keel size – this correspondsto an increase in size of the noncarnivorous keel of the pitcherand a decrease in size of the prey-catching tube. Thus, withincreasing nitrogen levels, the capacity for carnivory decreasedwhile maximum photosynthesis rates increased. In effect, the plantsswitched effort from nitrogen to carbon capture when morenitrogen was available in their environment.

9.5.3 The optimal foraging approach to diet width

Predators and prey have undoubtedlyinfluenced one another’s evolution.This can be seen in the distasteful or

poisonous leaves of many plants, in the spines of hedgehogs andin the camouflage coloration of many insect prey; and it can beseen in the stout ovipositors of wood wasps, the multichamberedstomachs of cattle and the silent approach and sensory excellenceof owls. Such specialization makes it clear, though, that no predatorcan possibly be capable of consuming all types of prey. Simpledesign constraints prevent shrews from eating owls (even thoughshrews are carnivores) and prevent humming-birds from eatingseeds.

Even within their constraints, however, most animals con-sume a narrower range of food types than they are morphologicallycapable of consuming. In trying to understand what determinesa consumer’s actual diet within its wide potential range, ecologistshave increasingly turned to optimal foraging theory. The aim of optimal foraging theory is to predict the foraging strategy to beexpected under specified conditions. It generally makes such pre-dictions on the basis of a number of assumptions:

1 The foraging behavior that isexhibited by present-day animals isthe one that has been favored by natural selection in the past butalso most enhances an animal’s fitness at present.

2 High fitness is achieved by a high net rate of energy intake(i.e. gross energy intake minus the energetic costs of obtain-ing that energy).

3 Experimental animals are observed in an environment to whichtheir foraging behavior is suited, i.e. it is a natural environmentvery similar to that in which they evolved, or an experimentalarena similar in essential respects to the natural environment.

••••

App

lied

N (

mg

l–1 )

0.01

0.2

1

Relative keel size

0.3 0.4 0.5 0.6 0.7 0.8 0.9

0.1

a plant that ‘switches’

diet width and

evolution

Figure 9.16 The relationship between relative keel size of pitchers of Sarracenia purpurea and nitrogen added as aerial spray in plots at Molly Bog, Vermont. Dotted lines indicate 95% confidenceintervals. A larger relative keel size corresponds to a reduced investment in organs of prey capture. (After Ellison & Gotelli, 2002.)

assumptions inherent

in optimal foraging

theory

EIPC09 10/24/05 2:01 PM Page 282


These assumptions will not always be justified. First, otheraspects of an organism’s behavior may influence fitness more thanoptimal foraging does. For example, there may be such a premiumon the avoidance of predators that animals forage at a place andtime where the risk from predators is lower, and in consequencegather their food less efficiently than is theoretically possible (see Section 9.5.4). Second, and just as important, for many consumers (particularly herbivores and omnivores) the efficientgathering of energy may be less critical than of some otherdietary constituent (e.g. nitrogen), or it may be of prime import-ance for the forager to consume a mixed and balanced diet. Insuch cases, the value of existing optimal foraging theory is limited. However, in circumstances where the energy maximiza-tion premise can be expected to apply, optimal foraging theoryoffers a powerful insight into the significance of the foraging ‘decisions’ that predators make (for reviews see Stephens & Krebs,1986; Krebs & Kacelnik, 1991; Sih & Christensen, 2001).

Typically, optimal foraging theorymakes predictions about foraging beha-vior based on mathematical modelsconstructed by ecological theoreticianswho are omniscient (‘all knowing’) asfar as their model systems are con-

cerned. The question therefore arises: is it necessary for a real forager to be equally omniscient and mathematical, if it is to adopt the appropriate, optimal strategy? The answer is ‘no’. Thetheory simply says that if there is a forager that in some way (inany way) manages to do the right thing in the right circumstances,then this forager will be favored by natural selection; and if itsabilities are inherited, these should spread, in evolutionary time,throughout the population.

Optimal foraging theory does not specify precisely how theforager should make the right decisions, and it does not requirethe forager to carry out the same calculations as the modeler. Laterwe consider another group of ‘mechanistic’ models (see Sec-tion 9.6.2) that attempt to show how a forager, given that it isnot omniscient, might nevertheless manage to respond by ‘rules of thumb’ to limited environmental information and therebyexhibit a strategy that is favored by natural selection. But it is optimal foraging theory that predicts the nature of the strategythat should be so favored.

The first paper on optimal foraging theory (MacArthur &Pianka, 1966) sought to understand the determination of diet ‘width’(the range of food types eaten by an animal) within a habitat.Subsequently, the model was developed into a more rigorous algebraic form, notably by Charnov (1976a). MacArthur andPianka argued that to obtain food, any predator must expend timeand energy, first in searching for its prey and then in handling it (i.e. pursuing, subduing and consuming it). Whilst searching,a predator is likely to encounter a wide variety of food items.MacArthur and Pianka therefore saw diet width as depending onthe responses of predators once they had encountered prey.

Generalists pursue (and may then subdue and consume) a largeproportion of the prey types they encounter; specialists continuesearching except when they encounter prey of their specificallypreferred type.

The ‘problem’ for any forager isthis: if it is a specialist, then it will onlypursue profitable prey items, but itmay expend a great deal of time and energy searching for them.Whereas if it is a generalist, it will spend relatively little time search-ing, but it will pursue both more and less profitable types of prey.An optimal forager should balance the pros and cons so as to max-imize its overall rate of energy intake. MacArthur and Piankaexpressed the problem as follows: given that a predator alreadyincludes a certain number of profitable items in its diet, shouldit expand its diet (and thereby decrease its search time) by includ-ing the next most profitable item as well?

We can refer to this ‘next most profitable’ item as the ith item.Ei/hi is then the profitability of the item, where Ei is its energycontent, and hi its handling time. In addition, K/M is the averageprofitability of the ‘present’ diet (i.e. one that includes all preytypes that are more profitable than i, but does not include preytype i itself ), and O is the average search time for the present diet.If a predator does pursue a prey item of type i, then its expectedrate of energy intake is Ei/hi. But if it ignores this prey item, whilstpursuing all those that are more profitable, then it can expect tosearch for a further O, following which its expected rate of energyintake is K/M. The total time spent in this latter case is O + M, andso the overall expected rate of energy intake is K/(O + M). The mostprofitable, optimal strategy for a predator will be to pursue theith item if, and only if:

Ei/hi ≥ K/(O + M). (9.1)

In other words, a predator should continue to add increasinglyless profitable items to its diet as long as Equation 9.1 is satisfied(i.e. as long as this increases its overall rate of energy intake). This will serve to maximize its overall rate of energy intake, K/(O + M).

This optimal diet model leads to a number of predictions.

1 Predators with handling times thatare typically short compared totheir search times should be gener-alists, because in the short time it takes them to handle a preyitem that has already been found, they can barely begin to searchfor another prey item. (In terms of Equation 9.1: Ei/hi is large(hi is small) for a wide range of prey types, whereas K/(O + M)is small (O is large) even for broad diets.) This predictionseems to be supported by the broad diets of many insectivo-rous birds feeding in trees and shrubs. Searching is always moderately time consuming, but handling the minute insectstakes negligible time and is almost always successful. A bird,

••••

theoreticians are

omniscient

mathematicians – the

foragers need not be

to pursue or not

pursue?

searchers should be

generalists

EIPC09 10/24/05 2:01 PM Page 283

284 CHAPTER 9

therefore, has something to gain and virtually nothing to loseby consuming an item once found, and overall profitability ismaximized by a broad diet.

2 By contrast, predators with hand-ling times that are long relative to their search times should be spe-cialists. That is, if O is always small,

then K/(O + M) is similar to K/M. Thus, maximizing K/(O + M) ismuch the same as maximizing K/h, which is achieved, clearly,by including only the most profitable items in the diet. Forinstance, lions live more or less constantly in sight of their preyso that search time is negligible; handling time, on the otherhand, and particularly pursuit time, can be long (and very energyconsuming). Lions consequently specialize on prey that canbe pursued most profitably: the immature, the lame and the old.

3 Other things being equal, a predatorshould have a broader diet in anunproductive environment (whereprey items are relatively rare and O is relatively large) than in a pro-ductive environment (where O is

smaller). This prediction is broadly supported by the twoexamples shown in Figure 9.17: in experimental arenas, bothbluegill sunfish (Lepomis macrochirus) and great tits (Parusmajor) had more specialized diets when prey density washigher. A related result has been reported from predators intheir natural setting – brown and black bears (Ursos arctos and

U. americanus) feeding on salmon in Bristol Bay in Alaska. Whensalmon availability was high, bears consumed less biomass percaptured fish, targeting energy-rich fish (those that had notspawned) or energy-rich body parts (eggs in females, brain inmales). In essence their diet became more specialized whenprey were abundant (Gende et al., 2001).

4 Equation 9.1 depends on the pro-fitability of the ith item (Ei/hi),depends on the profitabilities of theitems already in the diet (K/M) anddepends on the search times foritems already in the diet (O ) and thus on their abundance. Butit does not depend on the search time for the ith item, si. Inother words, predators should ignore insufficiently profitablefood types irrespective of their abundance. Re-examining theexamples in Figure 9.17, we can see that these both refer tocases in which the optimal diet model does indeed predict thatthe least profitable items should be ignored completely. Theforaging behavior was very similar to this prediction, but inboth cases the animals consistently took slightly more thanexpected of the less profitable food types. In fact, this sort ofdiscrepancy has been uncovered repeatedly, and there are anumber of reasons why it may occur, which can be summar-ized crudely by noting that the animals are not omniscient.The optimal diet model, however, does not predict a perfectcorrespondence between observation and expectation. It predicts the sort of strategy that will be favored by natural selection, and says that the animals that come closest to this

••••

handlers should be

specialists

specialization should

be greater in

productive

environments

the abundance of

unprofitable prey

types is irrelevant

Prediction ofoptimal diettheory

(a) Bluegill sunfish

Ratioencountered

Observedratio in diet

0.80 0.4Low density

0.80 0.4Medium density

0.80 0.4High density

SML

SML

Small preyMedium preyLarge prey

Predictedproportionin diet

(b) Great tit

Proportionencountered

Observedproportionin diet

0.80 0.4Low density

SL

SL

Small preyLarge prey

0.80 0.4High density I

0.80 0.4High density II

0.80 0.4High density III

Figure 9.17 Two studies of optimal diet choice that show a clear but limitedcorrespondence with the predictions ofCharnov’s (1976a) optimal diet model.Diets are more specialized at high preydensities; but more low profitability itemsare included than predicted by the theory.(a) Bluegill sunfish preying on different sizeclasses of Daphnia: the histograms showratios of encounter rates with each sizeclass at three different densities, togetherwith the predicted and observed ratios inthe diet. (After Werner & Hall, 1974.) (b) Great tits preying on large and smallpieces of mealworm. (After Krebs et al.,1977.) The histograms in this case refer to the proportions of the two types of item taken. (After Krebs, 1978.)

EIPC09 10/24/05 2:01 PM Page 284


strategy will be most favored. From this point of view, the cor-respondence between data and theory in Figure 9.17 seemsmuch more satisfactory. Sih and Christensen (2001) reviewed134 studies of optimal diet theory, focusing on the questionof what factors might explain the ability of the theory to correctly predict diets. Contrary to their a priori prediction, forager groups (invertebrate versus ectothermic vertebrateversus endothermic vertebrate) did not differ in the likelihoodof corroborating the theory. Their major conclusion was thatwhile optimal diet theory generally works well for foragers thatfeed on immobile or relatively immobile prey (leaves, seeds,mealworms, zooplankton relative to fish), it often fails to pre-dict diets of foragers that attack mobile prey (small mammals,fish, zooplankton relative to insect predators). This may be because variations among mobile prey in vulnerability(encounter rate and capture success) are often more import-ant in determining predator diets than are variations in the activechoices of predators (Sih & Christensen, 2001).

5 Equation 9.1 also provides a context for understanding the narrow specialization of predators that live in intimate asso-ciation with their prey, especially where an individual pre-dator is linked to an individual prey (e.g. many parasitoids andparasitic herbivores – and many parasites (see Chapter 12)).Since their whole lifestyle and life cycle are finely tuned to thoseof their prey (or host), handling time (M) is low; but this pre-cludes their being finely tuned to other prey species, forwhich, therefore, handling time is very high. Equation 9.1 willthus only apply within the specialist group, but not to any fooditem outside it.

On the other hand, polyphagy has definite advantages. Search costs(O) are typically low – food is easy to find – and an individual isunlikely to starve because of fluctuations in the abundance of onetype of food. In addition, polyphagous consumers can, of course,construct a balanced diet, and maintain this balance by varyingpreferences to suit altered circumstances, and can avoid consuminglarge quantities of a toxin produced by one of its food types. Theseare considerations ignored by Equation 9.1.

Overall, then, evolution maybroaden or restrict diets. Where preyexert evolutionary pressures demandingspecialized morphological or physio-logical responses from the consumer,

restriction is often taken to extremes. But where consumers feedon items that are individually inaccessible or unpredictable or lacking in certain nutrients, the diet often remains broad. An appeal-ing and much-discussed idea is that particular pairs of predatorand prey species have not only evolved but have coevolved. In other words, there has been an evolutionary ‘arms race’,whereby each improvement in predatory ability has been followedby an improvement in the prey’s ability to avoid or resist the predator, which has been followed by a further improvement in

predatory ability, and so on. This may itself be accompanied, on a long-term, evolutionary timescale, by speciation, so that, forexample, related species of butterfly are associated with relatedspecies of plants – all the species of the Heliconiini feed on mem-bers of the Passifloracaea (Ehrlich & Raven, 1964; Futuyma & May,1992). To the extent that coevolution occurs, it may certainly be an additional force in favor of diet restriction. At present, however, hard evidence for predator–prey or plant–herbivorecoevolution is proving difficult to come by (Futuyma & Slatkin,1983; Futuyma & May, 1992).

There may seem, at first sight, to be a contradiction betweenthe predictions of the optimal diet model and switching. In thelatter, a consumer switches from one prey type to another as theirrelative densities change. But the optimal diet model suggests thatthe more profitable prey type should always be taken, irrespect-ive of its density or the density of any alternative. Switching ispresumed to occur, however, in circumstances to which theoptimal diet model does not strictly apply. Specifically, switchingoften occurs when the different prey types occupy differentmicrohabitats, whereas the optimal diet model predicts behaviorwithin a microhabitat. Moreover, most other cases of switchinginvolve a change in the profitabilities of items of prey as their dens-ity changes, whereas in the optimal diet model these are constants.Indeed, in cases of switching, the more abundant prey type is themore profitable, and in such a case the optimal diet model predictsspecialization on whichever prey type is more profitable (that is,whichever is more abundant; in other words, switching).

9.5.4 Foraging in a broader context

It is worth stressing that foraging strat-egies will not always be strategies forsimply maximizing feeding efficiency.On the contrary, natural selection willfavor foragers that maximize their netbenefits, and strategies will thereforeoften be modified by other, conflicting demands on the indi-viduals concerned. In particular, the need to avoid predators will frequently affect an animal’s foraging behavior.

This has been shown in work on foraging by nymphs of an aquatic insect predator, the backswimmer Notonecta hoffmanni(Sih, 1982). These animals pass through five nymphal instars(with I being the smallest and youngest, and V the oldest), andin the laboratory the first three instars are liable to be preyed upon by adults of the same species, such that the relative risk ofpredation from adults was:

I > II > III > IV = V ≅ no risk.

These risks appear to modify the behavior of the nymphs, in thatthey tend (both in the laboratory and in the field) to avoid the

••••

coevolution:

predator–prey arms

races?

backswimmers forage

suboptimally but

avoid being preyed

on . . .

EIPC09 10/24/05 2:01 PM Page 285

286 CHAPTER 9

central areas of water bodies, where the concentration of adultsis greatest. In fact, the relative degree of avoidance was the sameas the relative risk of predation from adults:

I > II > III > IV = V ≅ no avoidance.

Yet these central areas also contain the greatest concentration of prey items for the nymphs, and so, by avoiding predators,nymphs of instars I and II showed a reduction in feeding rate in the presence of adults (although those of instar III did not). The young nymphs displayed a less than maximal feeding rate as a result of their avoidance of predation, but an increased survivorship.

The modifying influence of predatorson foraging behavior has also beenstudied by Werner et al. (1983b) work-

ing on bluegill sunfish. They estimated the net energy returns fromforaging in three contrasting laboratory habitats – in open water,amongst water weeds and on bare sediment – and they exam-ined how prey densities varied in comparable natural habitats ina lake through the seasons. They were then able to predict thetime at which the sunfish should switch between different lakehabitats so as to maximize their overall net energy returns. In the

absence of predators, three sizes of sunfish behaved as predicted(Figure 9.18). But in a further field experiment, this time in thepresence of predatory largemouth bass, the small sunfish restrictedtheir foraging to the water weed habitat (Figure 9.19) (Werner et al., 1983a). Here, they were relatively safe from predation,although they could only achieve a markedly submaximal rate ofenergy intake. By contrast, the larger sunfish are more or less safefrom predation by bass, and they continued to forage accordingto the optimal foraging predictions. In a similar vein, the nymphsof several species of algivorous mayflies largely restrict theirfeeding to the hours of darkness in streams that contain browntrout, reducing their overall feeding rates but also reducing therisk of predation (Townsend, 2003). In the case of mammals thatfeed at night, including mice, porcupines and hares, time spentfeeding may be reduced in bright moonlight when predation riskis highest (Kie, 1999).

A foraging strategy is an integralpart of an animal’s overall pattern ofbehavior. The strategy is stronglyinfluenced by the selective pressures favoring the maximizationof feeding efficiency, but it may also be influenced by other, pos-sibly conflicting demands. It is also worth pointing out one otherthing. The places where animals occur, where they are maximally

••••

Pre

dict

ed n

eten

ergy

gai

n (J

s–1

)

(a)

Per

cent

age

ofto

tal d

iet

15

Jul

0.2

0.8

0.6

0.4

0.2

0.8

0.6

0.4

Small Medium Large

31 15 31 15 30

Aug Sep

(b)

100

0.2

0.8

0.6

0.4

80

60

40

20

15

Jul

31 15 31 15 30

Aug Sep

100

80

60

40

20

15

Jul

31 15 31 15 30

Aug Sep

100

80

60

40

20

Open water

Sediments

Vegetation

Figure 9.18 Seasonal patterns in (a) the predicted habitat profitabilities (net rate of energy gain) and (b) the actual percentage of the diet originating from each habitat, for three size classes of bluegill sunfish (Lepomis macrochirus). Piscivores were absent. (The ‘vegetation’habitat is omitted from (b) for the sake of clarity – only 8–13% of the diet originated from this habitat for all size classes of fish.) There isgood correspondence between the patterns in (a) and (b). (After Werner et al., 1983b.)

. . . as do certain fish

predation and

the realized niche

EIPC09 10/24/05 2:01 PM Page 286


abundant and where they choose to feed are all key componentsof their ‘realized niches’. We saw in Chapter 8 that realizedniches can be highly constrained by competitors. Here, we seethat they can also be highly constrained by predators. This is also seen in the effects of predation by the barn owl (Tyto alba)on the foraging behavior of three heteromyid rodents, theArizona pocket mouse (Perognathus amplus), Bailey’s pocket mouse(P. baileyi) and Merriam’s kangaroo rat (Dipodomys merriami)(Brown et al., 1988). In the presence of owls, all three species moved to microhabitats where they were less at risk from owlpredation and where they reduced their foraging activity.However they did so to varying extents, such that the way in which the microhabitat was partitioned between them was quitedifferent in the presence and absence of owls.

9.6 Foraging in a patchy environment

For all consumers, food is distributedpatchily. The patches may be naturaland discrete physical objects: a bush

laden with berries is a patch for a fruit-eating bird; a leaf coveredwith aphids is a patch for a predatory ladybird. Alternatively, a‘patch’ may only exist as an arbitrarily defined area in an appar-ently uniform environment; for a wading bird feeding on a sandybeach, different 10 m2 areas may be thought of as patches thatcontain different densities of worms. In all cases though, a patchmust be defined with a particular consumer in mind. One leaf isan appropriate patch for a ladybird, but for a larger and more activeinsectivorous bird, 1 m2 of canopy or even a whole tree may represent a more appropriate patch.

Ecologists have been particularly interested in patch preferencesof consumers where patches vary in the density of food or preyitems they contain. There are many examples where predatorsshow an ‘aggregative response’, spending more time in patchescontaining high densities (because these are the most profitablepatches) (Figure 9.20a–d), although such direct density dependenceis not always the case (Figure 9.20e). We deal with aggregativeresponses in more detail in Chapter 10 where their importancein population dynamics will be our focus, and particularly theirpotential to lend stability to predator–prey dynamics. For now,we concentrate on the behavior that leads to predator aggrega-tion (Section 9.6.1), the optimal foraging approach to patch use(Section 9.6.2) and the distribution patterns that are likely to resultwhen the opposing tendencies of predators to aggregate and tointerfere with each other’s foraging are both taken into account(Section 9.6.3).

9.6.1 Behavior that leads to aggregated distributions

There are various types of behaviorunderlying the aggregative responsesof consumers, but they fall into two broad categories: thoseinvolved with the location of profitable patches, and theresponses of consumers once within a patch. The first categoryincludes all examples in which consumers perceive, at a distance,the existence of heterogeneity in the distribution of their prey.

Within the second category –responses of consumers within patches– there are two main aspects of behav-ior. The first is a change in the consumer’s pattern of searchingafter encountering items of food. In particular, there is often aslowing down of movement and an increased rate of turning imme-diately following the intake of food, both of which lead to theconsumer remaining in the vicinity of its last food item (‘area-restricted search’). Alternatively, or in addition, consumers maysimply abandon unprofitable patches more rapidly than theyabandon profitable ones. Both types of behavior were evident whenthe carnivorous, net-spinning larva of the caddis-fly Plectrocnemiaconspersa feeds on chironomid (midge) larvae in a laboratorystream. Caddis in their nets were provided with one prey itemat the beginning of the experiment and then fed daily rations of

••••

Num

ber

of fi

sh25

(a) Predator present

20

15

10

5

10060200 40

Percentage vegetation where sunfishprey are taken from

80

Num

ber

of fi

sh

0

50

(b) No predator present

45

40

35

30

25

20

15

10

5

Figure 9.19 (a) In contrast to Figure 9.18 and to (b), whenlargemouth bass (which prey on small bluegill sunfish) are present many sunfish take prey from areas where the percentagevegetation is high and where they are relatively protected frompredation. (After Werner et al., 1983a.)

food is patchily

distributed

locating a patch

area-restricted search

EIPC09 10/24/05 2:01 PM Page 287

•• ••

288 CHAPTER 9

30100

0

10

30

40

20

Aphids per leaf

(a)

20

Sea

rchi

ng ti

me

per

leaf

(m

in)

40000

0

10

2000

Number of Corophium (m–2)

(b)

5

Num

ber

of r

edsh

ank

(ha–

1 )

Par

asiti

sm (

%)

25155

0

0

20

40

50

10

Host density per patch

(c)

10

30

20 10006002000

0

20

40

60

400


(d)

10

30

800

50

8006002000

0

20

60

100

400


(e)

40

80

zero, one or three prey. The tendency to abandon the net waslowest at the higher feeding rates (Townsend & Hildrew, 1980).Plectrocnemia’s behavior in relation to prey patches also has an element of area-restricted search: the likelihood that it will spin a net in the first place depends on whether it happens toencounter a food item (which it can consume even without a net) (Figure 9.21a). Overall, therefore, a net is more likely to beconstructed, and less likely to be abandoned, in a rich patch. These two behaviors account for a directly density-dependentaggregative response in the natural stream environment observedfor much of the year (Figure 9.21b).

The difference in the rates of aban-donment of patches of high and lowprofitability can be achieved in a num-ber of ways, but two are especially

easy to envisage. A consumer might leave a patch when its feed-ing rate drops below a threshold level, or a consumer might havea giving-up time – it might abandon a patch whenever a particu-lar time interval passes without the successful capture of food.Whichever mechanism is used, or indeed if the consumer simplyuses area-restricted search, the consequences will be the same:

individuals will spend longer in more profitable patches, andthese patches will therefore generally contain more consumers.

9.6.2 Optimal foraging approach to patch use

The advantages to a consumer of spending more time in higherprofitability patches are easy to see. However, the detailed alloca-tion of time to different patches is a subtle problem, since it depends on the precise differentials in profitability, the averageprofitability of the environment as a whole, the distance betweenthe patches, and so on. The problem has been a particular focusof attention for optimal foraging theory. In particular, a great dealof interest has been directed at the very common situation in whichforagers themselves deplete the resources of a patch, causing itsprofitability to decline with time. Amongst the many examplesof this are insectivorous insects removing prey from a leaf, andbees consuming nectar from a flower.

Charnov (1976b) and Parker and Stuart (1976) produced similar models to predict the behavior of an optimal forager insuch situations. They found that the optimal stay-time in a patch

Figure 9.20 Aggregative responses: (a) coccinellid larvae (Coccinella septempunctata) spend more time on leaves with high densities of their aphid prey (Brevicoryne brassicae) (after Hassell & May, 1974); (b) redshank (Tringa totanus) aggregate in patches with higher densitiesof their amphipod prey (Corophium volutator) (after Goss-Custard, 1970); (c) direct density dependence when the parasitoid Delia radicumattacks Trybliographa rapae; and (d) direct density dependence when the parasitoid Aspidiotiphagus citrinus attacks Fiorinia externa. (e) Butdirect density dependence is not always the case: inverse density dependence when the parasitoid Ooencyrtus kuwanai attacks Lymantriadispar. ((c–e) after Pacala & Hassall, 1991.)

thresholds and

giving-up times

EIPC09 10/24/05 2:01 PM Page 288

••


should be defined in terms of the rate of energy extraction experienced by the forager at the moment it leaves a patch (the ‘marginal value’ of the patch). Charnov called the results the‘marginal value theorem’. The models were formulated mathe-matically, but their salient features are shown in graphic form inFigure 9.22.

The primary assumption of the model is that an optimal forager will maximize its overall intake of a resource (usuallyenergy) during a bout of foraging, taken as a whole. Energy will,in fact, be extracted in bursts because the food is distributed patchily;the forager will sometimes move between patches, during whichtime its intake of energy will be zero. But once in a patch, theforager will extract energy in a manner described by the curvesin Figure 9.22a. Its initial rate of extraction will be high, but astime progresses and the resources are depleted, the rate ofextraction will steadily decline. Of course, the rate will itselfdepend on the initial contents of the patch and on the forager’sefficiency and motivation (Figure 9.22a).

The problem under considerationis this: at what point should a foragerleave a patch? If it left all patchesimmediately after reaching them, thenit would spend most of its time travel-

ing between patches, and its overall rate of intake would be low.If it stayed in all patches for considerable lengths of time, then itwould spend little time traveling, but it would spend extendedperiods in depleted patches, and its overall rate of intake wouldagain be low. Some intermediate stay-time is therefore optimal.In addition, though, the optimal stay-time must clearly be greaterfor profitable patches than for unprofitable ones, and it must dependon the profitability of the environment as a whole.

Consider, in particular, the forager in Figure 9.22b. It is for-aging in an environment where food is distributed patchily and

where some patches are more valuable than others. The averagetraveling time between patches is tt. This is therefore the lengthof time the forager can expect to spend on average after leavingone patch before it finds another. The forager in Figure 9.22b hasarrived at an average patch for its particular environment, and ittherefore follows an average extraction curve. In order to forageoptimally it must maximize its rate of energy intake not merelyfor its period in the patch, but for the whole period since its depar-ture from the last patch (i.e. for the period tt + s, where s is thestay-time in the patch).

If it leaves the patch rapidly then this period will be short (tt + sshort in Figure 9.22b). But by the same token, little energywill be extracted (Eshort). The rate of extraction (for the whole periodtt + s) will be given by the slope of the line OS (i.e. Eshort/(tt + sshort)).On the other hand, if the forager remains for a long period (slong)then far more energy will be extracted (Elong); but, the overall rateof extraction (the slope of OL) will be little changed. To maxim-ize the rate of extraction over the period tt + s, it is necessary tomaximize the slope of the line from O to the extraction curve.This is achieved simply by making the line a tangent to the curve(OP in Figure 9.22b). No line from O to the curve can be steeper,and the stay-time associated with it is therefore optimal (sopt).

The optimal solution for the for-ager in Figure 9.22b, therefore, is toleave that patch when its extractionrate is equal to (tangential to) the slopeof OP, i.e. it should leave at point P. In fact, Charnov, and Parkerand Stuart, found that the optimal solution for the forager is toleave all patches, irrespective of their profitability, at the sameextraction rate (i.e. the same ‘marginal value’). This extraction rateis given by the slope of the tangent to the average extraction curve(e.g. in Figure 9.22b), and it is therefore the maximum averageoverall rate for that environment as a whole.

••

30 min 30 min

30 min 30 min

0

1.0

0.5

Pro

port

ion

of la

rvae

(a)

1 2

Fed (n = 38)Widespread movement

0

1.0

0.5

1 2

Unfed (n = 42)

0

1.0

0.5

1 2

Net-building

0

1.0

0.5

1 2Time (h)

0

4

2

3

1

0 10 20 30 50 +

Mea

n nu

mbe

r of

pred

ator

s pe

r sa

mpl

e

(b)

Biomass of preyper sample (mg dry weight)

Figure 9.21 (a) On arrival in a patch,fifth-instar Plectrocnemia conspersa larvaethat encounter and eat a chironomid preyitem at the beginning of the experiment(‘fed’) quickly cease wandering andcommence net-building. Predators that failto encounter a prey item (‘unfed’) exhibitmuch more widespread movement duringthe first 30 min of the experiment, and aresignificantly more likely to move out of the patch. (b) Directly density-dependentaggregative response of fifth-instar larvae ina natural environment expressed as meannumber of predators against combinedbiomass of chironomid and stonefly prey per 0.0625 m2 sample of streambed (n = 40). (After Hildrew & Townsend,1980; Townsend & Hildrew, 1980.)

when should a

forager leave a patch

that it is depleting?

how to maximize

overall energy intake

EIPC09 10/24/05 2:01 PM Page 289

•• ••

290 CHAPTER 9

TimeEnters patch

Rat

e of

ene

rgy

extr

actio

n (

)or

cum

ulat

ive

ener

gy e

xtra

cted

(

)

High productivity patch and/or highforager efficiency

High productivity patch and/or high forager efficiencyLow productivity patch and/or high forager efficiency

Low productivity patch and/or highforager efficiency

0

(a)

Cum

ulat

ive

ener

gy e

xtra

cted

TimeO

S

P

L

(b)

Elong

Eopt

Eshort

tt tt + slongtt + sshort tt + sopt

Highproductivitypatch

Lowproductivitypatch

Average

Cum

ulat

ive

ener

gy e

xtra

cted

(c)

shighslong tt

sshort ttslow

(d)

Short tt

Highaverage

Low averageLong tt

slowshigh

(e)

Figure 9.22 The marginal value theorem. (a) When a forager enters a patch, its rate of energy extraction is initially high (especially in a highly productive patch or where the forager has a high foraging efficiency), but this rate declines with time as the patch becomesdepleted. The cumulative energy intake approaches an asymptote. (b) The options for a forager. The solid colored curve is cumulativeenergy extracted from an average patch, and tt is the average traveling time between patches. The rate of energy extraction (which shouldbe maximized) is energy extracted divided by total time, i.e. the slope of a straight line from the origin to the curve. Short stays in thepatch (slope = Eshort/(tt + sshort)) and long stays (slope = Elong/(tt + slong)) both have lower rates of energy extraction (shallower slopes) than astay (sopt) which leads to a line just tangential to the curve. sopt is therefore the optimum stay-time, giving the maximum overall rate ofenergy extraction. All patches should be abandoned at the same rate of energy extraction (the slope of the line OP). (c) Low productivitypatches should be abandoned after shorter stays than high productivity patches. (d) Patches should be abandoned more quickly whentraveling time is short than when it is long. (e) Patches should be abandoned more quickly when the average overall productivity is highthan when it is low.

EIPC09 10/24/05 2:01 PM Page 290

••


The model therefore confirms thatthe optimal stay-time should be greaterin more productive patches than inless productive patches (Figure 9.22c).

Moreover, for the least productive patches (where the extractionrate is never as high as OP) the stay-time should be zero. Themodel also predicts that all patches should be depleted such thatthe final extraction rate from each is the same (i.e. the ‘marginalvalue’ of each is the same); and it predicts that stay-times shouldbe longer in environments where the traveling time betweenpatches is longer (Figure 9.22d) and that stay-times should be longerwhere the environment as a whole is less profitable (Figure 9.22e).

Encouragingly, there is evidencefrom a number of cases that lends sup-port to the marginal value theorem. Inone of the first tests of the theory,

Cowie (1977) considered the prediction set out in Figure 9.22d:that a forager should spend longer in each patch when the travel-ing time is longer. He used captive great tits in a large indooraviary, and got the birds to forage for small pieces of mealwormhidden in sawdust-filled plastic cups – the cups were ‘patches’. Allpatches on all occasions contained the same number of prey, but traveling time was manipulated by covering the cups withcardboard lids that varied in their tightness and therefore variedin the time needed to prize them off. Birds foraged alone, andCowie used six in all, subjecting each to two habitats. One of thesehabitats always had longer traveling times (tighter lids) than the other. For each bird in each habitat Cowie measured the average traveling time and the curve of cumulative food intakewithin a patch. He then used the marginal value theorem to predict the optimal stay-time in habitats with different traveling

times, and compared these predictions with the stay-times he actually observed. As Figure 9.23 shows, the correspondencewas quite close. It was closer still when he took account of the fact that there was a net loss of energy when the birds weretraveling between patches.

Predictions of the marginal value theorem have also been exam-ined through the behavior of the egg parasitoid, Anaphes victus,attacking the beetle Listronotus oregonensis in a laboratory setting(Boivin et al., 2004). Patches differed in quality by virtue of thevarying proportions of hosts already parasitized at the start of theexperiment, and in line with the theorem’s predictions, parasitoidsstayed longer in the more profitable patches (Figure 9.24a).However, contrary to a further prediction, the marginal rate offitness gain (the rate of progeny production in the final 10 minbefore leaving a patch) was greatest in the initially mostprofitable patches (Figure 9.24b).

As was the case with optimal diettheory, the risk of being preyed uponcan be expected to modify the pre-dicted outcomes of optimal patch use. With this in mind, Morris andDavidson (2000) compared the giving-up food extraction rates of white-footed mice (Peromyscus leucopus) in a forest habitat (where predation risk is low) and aforest-edge habitat (where predation risk is high). They provided‘patches’ (containers) with 4 g of millet grain in 11 foraging sitesin the two habitat types, and in both habitat types some sites werein relatively open situations and others were beneath shrubs. Theythen monitored the grain remaining at the time that the patcheswere abandoned on two separate days. Their results (Figure 9.25)supported the predictions that mice should abandon patches at

••

Travel time (s)

(a) (c)

0

96

8 3216 24

80

64

48

32

16

Tim

e in

pat

ch (

s)

(b)

0

96

8 3216 24

80

64

48

32

16

Figure 9.23 (a) An experimental ‘tree’ for great tits, with three patches. (b) Predicted optimal time in a patch plotted against travelingtime ( ), together with the observed mean points (± SE) for six birds, each in two environments. (c) The same data points, and thepredicted time taking into account the energetic costs of traveling between patches. (After Cowie, 1977; from Krebs, 1978.)

predictions of

the marginal

value theorem . . .

. . . supported by

some experiments

optimal patch

use predictions are

modified when there

is a risk of being

preyed upon

EIPC09 10/24/05 2:01 PM Page 291

292 CHAPTER 9

higher harvest rates in vulnerable edge habitats than in safe forest habitats, particularly in open situations (where predationrisk is highest in each habitat).

A much fuller review of tests of themarginal value theorem is provided,for example, by Krebs and Kacelnik(1991). The picture this conveys is one of encouraging but not perfectcorrespondence – much like the balance

of the results presented here. The main reason for the imperfec-tion is that the animals, unlike the modelers, are not omniscient.As was clear in the case of the white-footed mice, they may need to spend time doing things other than foraging (e.g. avoiding predators). Foragers may also need to spend time learning about and sampling their environment, and are none the less likely toproceed in their foraging with imperfect information about the

distribution of their hosts. For the parasitoids in Figure 9.24, forexample, Boivin et al. (2004) suggest that they seem to base theirassessment of overall habitat quality on the quality of the first patchthey encounter; that is, they ‘learn’ but their learned assessmentmay still be wrong. Such a strategy would be adaptive, though,if there was considerable variation in quality between generations(so that each generation had to learn anew), but little variationin quality between patches within a generation (so that the firstpatch encountered was a fair indication of quality overall).

Nevertheless, in spite of their limited information, animals seem often to come remarkably close to the predicted strategy.Ollason (1980) developed a mechanistic model to account for thisin the great tits studied by Cowie. Ollason’s is a memory model.It assumes that an animal has a ‘remembrance of past food’, whichOllason likens to a bath of water without a plug. Fresh remem-brance flows in every time the animal feeds. But remembranceis also draining away continuously. The rate of input depends on the animal’s feeding efficiency and the productivity of the current feeding area. The rate of outflow depends on the animal’sability to memorize and the amount of remembrance. Remem-brance drains away quickly, for example, when the amount is large(high water level) or the memorizing ability is poor (tall, narrowbath). Ollason’s model simply proposes that an animal should stayin a patch until remembrance ceases to rise; an animal should leavea patch when its rate of input from feeding is slower than its rateof declining remembrance.

An animal foraging consistentlywith Ollason’s model behaves in a wayvery similar to that predicted by themarginal value theorem. This is shown for the case of Cowie’sgreat tits in Figure 9.26. As Ollason himself remarks, this showsthat to forage in a patchy environment in a way that approximatesclosely to optimality, an animal need not be omniscient, it doesnot need to sample and it does not need to perform numericalanalyses to find the maxima of functions of many variables: all it

••••

2000Pat

ch r

esid

ence

tim

e (s

)

10075250

0

8000

50

Initial percentage of parasitized hosts

(a)

7000

6000

5000

4000

3000

1000

48 h

3 h

Pro

geny

pro

duce

d (n

o. m

in–1

)

10075250.000

0

0.003

50

Initial percentage of parasitized hosts

(b)

0.002

0.001

48 h

3 h

Giv

ing-

up d

ensi

ty (

g)

10

0.5

2.5

3.5

3.0

1.0

1.5

2.0

2 1 2Day

Edge

Forest

Shrub

Open

Figure 9.24 (a) When the parasitoidAnaphes victus attacked the beetleListronotus oregonensis in patches of 16 hosts, a varying percentage of whichhad already been parasitized, parasitoidsremained longer in the more profitablepatches: those with the smaller percentageof parasitized hosts. (b) However, themarginal gain rate in fitness – the numberof progeny produced per minute in thefinal 10 min before leaving a patch – wasgreatest in the initially most profitablepatches. Bars represent standard errors.(After Boivin et al., 2004.)

Figure 9.25 The mass of millet grain remaining (giving-updensity, g) was higher in patches in the open (riskier) than inpaired patches located under shrubs (safer), and was higher inforest-edge habitat (higher predation) than in forest (lowerpredation). (After Morris & Davidson, 2000.)

predicted and

observed behaviors

do not correspond

perfectly mechanistic models

of optimal foraging

EIPC09 10/24/05 2:01 PM Page 292


needs to do is to remember, and to leave each patch if it is notfeeding as fast as it remembers doing. As Krebs and Davies(1993) point out, this is no more surprising than the observationthat the same birds can fly without any formal qualification inaerodynamics.

Mechanistic models have also been developed and tested fora range of patterns of parasitoid attack (like that in Figure 9.24)(see Vos et al., 1998; Boivin et al., 2004). These highlight the import-ant distinction between ‘rule of thumb’ behavior, where animalsfollow innate and unvarying rules, and learned behavior, whererules are subject to modification in the light of the forager’simmediate experience. The weight of evidence suggests thatlearning plays at least some role in most foragers’ decisions.There is an important distinction, too, between ‘incremental’ and‘decremental’ behavior. With incremental behavior each suc-cessful attack in a patch increases the forager’s chance of stayingthere. This is likely to be adaptive when there is considerable variation in quality between patches, because it encourageslonger stay-times in better quality patches. With ‘decremental’behavior each successful attack in a patch decreases the forager’schance of staying there. This is likely to be adaptive when all patchesare of approximately the same quality, because it encourages foragers to leave depleted patches.

Thus, Ollason’s model for great tits incorporated rule ofthumb, incremental behavior. Boivin et al., on the other hand,found their parasitoids to be exhibiting learned, decrementalbehavior: a parasitoid attacking a healthy host, for example, was1.43 times more likely to leave a patch thereafter, and one reject-ing a host that had already been attacked was 1.11 times morelikely to leave. Vos et al. (1998), by contrast, found incrementalbehavior when the parasitoid Cotesia glomerata attacked its

butterfly larva host, Pieris brassicae: each successful encounterincreased its tendency to remain in a patch. For both the greattit and parasitoids, therefore, optimal foraging and mechanisticmodels are seen to be compatible and complementary in explain-ing how a predator has achieved its observed foraging pattern,and why that pattern has been favored by natural selection.

Finally, the principles of optimalforaging are also being applied toinvestigations of the foraging strategiesof plants for nutrients (reviewed byHutchings & de Kroon, 1994). When does it pay to produce longstolons moving rapidly from patch to patch? When does it payto concentrate root growth within a limited volume, foraging froma patch until it is close to depletion? Certainly, it is good to seesuch intellectual cross-fertilization across the taxonomic divide.

9.6.3 Ideal free and related distributions: aggregation and interference

We can see, then, that consumers tendto aggregate in profitable patches wheretheir expected rate of food consumptionis highest. Yet we might also expect that consumers will competeand interfere with one another (discussed further in Chapter 10),thereby reducing their per capita consumption rate. It follows from this that patches that are initially most profitable becomeimmediately less profitable because they attract most consumers.We might therefore expect the consumers to redistribute them-selves, and it is perhaps not surprising that the observed patternsof predator distributions across prey patches vary substantially fromcase to case. But can we make some sense of this variation in pattern?

In an early attempt to do so, it wasproposed that if a consumer foragesoptimally, the process of redistributionwill continue until the profitabilities ofall patches are equal (Fretwell & Lucas, 1970; Parker, 1970). Thiswill happen because as long as there are dissimilar profitabilities,consumers should leave less profitable patches and be attractedto more profitable ones. Fretwell and Lucas called the consequentdistribution the ideal free distribution: the consumers are ‘ideal’in their judgement of profitability, and ‘free’ to move from patchto patch. Consumers were also assumed to be equal. Hence, withan ideal free distribution, because all patches come to have thesame profitability, all consumers have the same consumption rate.There are some simple cases where consumers appear to conformto an ideal free distribution insofar as they distribute themselvesin proportion to the profitabilities of different patches (e.g.Figure 9.27a), but even in such cases one of the underlyingassumptions is likely to have been violated (e.g. Figure 9.27b –all consumers are not equal).

••••

Tim

e in

pat

ch (

s)

322480

0

16

48

96

16

Travel time (s)

32

64

80

Figure 9.26 Cowie’s (1977) great tit data (see Figure 9.23)compared to the predictions of Ollason’s (1980) mechanisticmemory model.

optimal foraging

in plants

the ideal free

distribution . . .

. . . is a balance

between attractive

and repellant forces

EIPC09 10/24/05 2:01 PM Page 293

294 CHAPTER 9

The early ideas have been muchmodified taking account, for example,of unequal competitors (see Milinski & Parker, 1991; Tregenza, 1995, for

reviews). In particular, ideal free theory was put in a more ecological context by Sutherland (1983) when he explicitly incorporated predator handling times and mutual interferenceamongst the predators. He found that predators should be dis-tributed such that the proportion of predators in site i, pi, is relatedto the proportion of prey (or hosts) in site i, hi, by the equation:

pi = k (hi1/m) (9.2)

where m is the coefficient of interference, and k is a ‘normaliz-ing constant’ such that the proportions, pi, add up to 1. It is nowpossible to see how the patch to patch distribution of predatorsmight be determined jointly by interference and the selection bythe predators of intrinsically profitable patches.

If there is no interference amongst the predators, then m = 0.All should exploit only the patch with the highest prey density(Figure 9.28), leaving lower density patches devoid of predators.

If there is a small or moderate amount of interference (i.e. m > 0, but m < 1 – a biologically realistic range), then high-densityprey patches should still attract a disproportionate number of predators (Figure 9.28). In other words, there should be an aggre-gative response by the predators, which is not only directly density dependent, but actually accelerates with increasing preydensity in a patch. Hence, the prey’s risk of predation might itselfbe expected to be directly density dependent: the greatest risk ofpredation in the highest prey density patches (like the examplesin Figure 9.20c, d).

With a little more interference (m ≈ 1) the proportion of thepredator population in a patch should still increase with the pro-portion of prey, but now it should do so more or less linearlyrather than accelerating, such that the ratio of predators : prey isroughly the same in all patches (Figure 9.28, and, for example,

Figure 9.20a). Here, therefore, the risk of predation might beexpected to be the same in all patches, and hence independentof prey density.

Finally, with a great deal of interference (m > 1) the highestdensity prey patches should have the lowest ratio of predators :prey (Figure 9.28). The risk of predation might therefore beexpected to be greatest in the lowest prey density patches, andhence inversely density dependent (like the data in Figure 9.20e).

It is clear, therefore, that the range of patterns amongst thedata in Figure 9.20 reflects a shifting balance between the forcesof attraction and of repulsion. Predators are attracted to highlyprofitable patches; but they are repelled by the presence of otherpredators that have been attracted in the same way.

This description, however, of therelationship between the distributionof predators and the distribution ofpredation risk has been peppered with ‘might be expected to’s. The

••••

Num

ber

of p

iece

s of

bre

ad e

aten

N

2

A

6

12

Individuals ranked in order of success

(b)

4

10

8

B C D E F G H I MJ K L

Num

ber

of d

ucks

4203001200

0

6

12

18

180

Time (s)

(a)

4

10

16

2

8

14

60 360240

Figure 9.27 (a) When 33 ducks were fed pieces of bread at two stations arounda pond (with a profitability ratio of 2 : 1), the number of ducks at the poorer station,shown here, rapidly approached one-thirdof the total, in apparent conformity withthe predictions of ideal free theory. (b)However, contrary to the assumptions and other predictions of simple theory, theducks were not all equal. (After Harper,1982; from Milinski & Parker, 1991.)

Pro

port

ion

of p

reda

tors

(p i

)

Proportion of prey in the i th patch (hi)

m = 0.3

m = 0.6

m = 1.0

m > 1

m 0

Figure 9.28 The effect of the interference coefficient, m, on theexpected distribution of predators amongst patches of prey varyingin the proportion of the total prey population they contain (andhence, in their ‘intrinsic’ profitability). (After Sutherland, 1983.)

incorporating a range

of interference

coefficients

pseudo-interference

EIPC09 10/24/05 2:01 PM 4


reason is that the relationship also depends on a range of factorsnot so far considered. For example, Figure 9.29 shows a case wherethe parasitoid Trichogramma pretiosum aggregates in high-densitypatches of its moth host, but the risk of parasitism to the mothis greatest in low-density host patches. The explanation probablylies in time wasted by parasitoids in high-density host patches, dealing with already parasitized hosts that may still attract para-sitoids because they are not physically removed from a patch (unlikepreyed-upon prey) (Morrison & Strong, 1981; Hassell, 1982). Thus,earlier parasitoids in a patch may interfere indirectly with laterarrivals, in that the previous presence of a parasitoid in a patchmay reduce the effective rate at which later arrivals attack unpara-sitized hosts. This effect has been termed ‘pseudo-interference’(Free et al., 1977); its potentially important effects on populationdynamics are discussed in Chapter 10.

Expected patterns are modified further still if we incorporate learningby the predators, or the costs of migra-tion between patches (Bernstein et al.,

1988, 1991). With a realistic value of m (= 0.3), the aggregativeresponse of predators is directly density dependent (as expected)as long as the predators’ learning response is strong relative tothe rate at which they can deplete patches. But if their learningresponse is weak, predators may be unable to track the changesin prey density that result from patch depletion. Their distributionwill then drift to one that is independent of the density of prey.

Similarly, when the cost of migration is low, the predators’aggregative response remains directly density dependent (with m = 0.3) (Figure 9.30a). When the cost of migration is increased,however, it still pays predators in the poorest patches to move,but for others the costs of migration can outweigh the potentialgains of moving. For these, the distribution across prey patches

is random. This results in inverse density dependence in mortalityrate between intermediate and good patches, and in a ‘domed’relationship overall (Figure 9.30b). When the cost of migrationis very high, it does not pay predators to move whatever patchthey are in – mortality is inversely density dependent across allpatches (Figure 9.30c).

Clearly, there is no shortage of potential causes for the widerange of types of distributions of predators, and of mortalityrates, across prey patches (see Figures 9.20 and 9.29). Their consequences, in terms of population dynamics, are one of thetopics dealt with in the chapter that follows. This highlights the crucial importance of forging links between behavioral and population ecology.

Summary

Predation is the consumption of one organism by another, in whichthe prey is alive when the predator first attacks it. There are two main ways in which predators can be classified. The first is‘taxonomic’ – carnivores consume animals, herbivores consumeplants, etc.– and the second is ‘functional’, in which true predators,grazers, parasitoids and parasites are distinguished.

The effects of herbivory on a plant depend on which herbi-vores are involved, which plant parts are affected, and the timing of attack relative to the plant’s development. Leaf-biting,sap-sucking, mining, flower and fruit damage and root pruningcan be expected to differ in the effect they have on the plant.Because the plant usually remains alive in the short term, the effectsof herbivory are also crucially dependent on the response of theplant. The evolutionary selection pressure exerted by herbivoreshas led to a variety of plant physical and chemical defenses that

••••

Cou

nts

per

patc

h

025

5

15

0

(a)

10

7550

100

Hosts per patch 2 4 8 16 32

Total parasitoids

Par

asiti

sm (

%)

025

20

80

1

(b)

60

7550

100

Hosts per patch 2 4 8 16 32

Total parasitoids

40

Fgiure 9.29 (a) The aggregative response of the egg parasitoid, Trichogramma pretiosum, which aggregates on patches with high densitiesof its host Plodia interpunctella. (b) The resultant distribution of ill effects: hosts on high-density patches are least likely to be parasitized.(After Hassell, 1982.)

learning and

migration

EIPC09 10/24/05 2:01 PM Page 295

296 CHAPTER 9

resist attack. These may be present and effective continuously (constitutive defense) or increased production may be induced byattack (inducible defense). It is not straightforward to determinewhether the supposed ‘defenses’ actually have measurable, neg-ative effects on the herbivore and positive consequences for theplant, especially after the costs of mounting the response havebeen taken into account. We discuss the difficulties of revealingsuch effects and review the relationships between herbivory andplant survival and fecundity.

More generally, the immediate effect of predation on a popu-lation of prey is not always predictably harmful, first because the individuals that are killed are not always a random sample(and may be those with the lowest potential to contribute to the population’s future) and second because of compensatorychanges in the growth, survival or reproduction of the survivingprey (especially via reduced competition for a limiting resource).From the predator’s point of view, an increase in the amount offood consumed can be expected to lead to increased rates of growth,development and birth, and decreased rates of mortality. How-ever, there are a number of factors that complicate this simplerelationship between consumption rate and consumer benefit.

Consumers can be classified on a continuum from monophagy(feeding on a single prey type) to polyphagy (many prey types).The preferences of many consumers are fixed – they are main-

tained irrespective of the relative availabilities of alternative foodtypes. But many others switch their preference, such that fooditems are eaten disproportionately often when they are common.A mixed diet may be favored first because each food type containsa different undesirable toxic chemical. More generally, a general-ist strategy would be favored if a consumer has more to gain thanlose in accepting low-quality items, once encountered, rather thanignoring them and continuing to search. We discuss this in thecontext of optimal diet theory, the aim of which is to predict theforaging strategy to be expected under specified conditions.

Food is generally distributed patchily and ecologists havebeen particularly interested in patch preferences of consumerswhere patches vary in the density of food or prey items they contain. We describe the behaviors that lead to aggregated dis-tributions and the nature of the distribution patterns that result.The advantages to a consumer of spending more time in higherprofitability patches are easy to see. However, the detailed alloca-tion of time to different patches is a subtle problem, dependingon the precise differentials in profitability, the average profitabil-ity of the environment as a whole, the distance between patches,and so on. This is the domain of the theory of optimal patch use.The predictions of both optimal foraging and optimal patch usetheory have to be modified when there is a simultaneous risk ofa consumer being preyed upon.

••

Mor

talit

y (a

rcsi

n tr

ansf

orm

ed)

200150500.5

0

0.7

1.0

1.1

100

Prey density

(a)

0.9

0.8

0.6

200150500.4

0

0.6

0.9

1.1

100

Prey density

(b)

0.8

0.7

0.5

1.0

200150500.2

0

1.2

100

Prey density

(c)

0.8

0.6

1.0

0.4

Figure 9.30 The effect of a cost to migration in predators distributing themselves across prey patches in a simulation model. Theinterference coefficient, m, is 0.3 and would lead to direct density dependence in the absence of a migration cost. (a) Low migration cost:direct density dependence is maintained. (b) Intermediate cost: a ‘domed’ relationship. (c) High cost: inverse density dependence. (AfterBernstein et al., 1991.)

EIPC09 10/24/05 2:01 PM Page 296

Chapter 9 The Nature of Predation Cap 9.pdf · rate at which the adult females lay eggs. Each egg...

Documents

Transcript of Chapter 9 The Nature of Predation Cap 9.pdf · rate at which the adult females lay eggs. Each egg...